by Tsutomu Shimomura & John Markoff    

(PDF Version)



take*down (tak/ doun/) adj. Sports. A move or maneuver in
wrestling or the martial arts in which a standing opponent is
forced to the floor.

The American Heritage Dictionary of the English Language,
Third Edition


If you find three men sitting alone in a van in a shopping mall
parking lot at two in the morning and one of them is holding an
odd looking antenna, there's usually only one conclusion you

   They're cops.

   I wasn't, and despite all the media frenzy that would erupt
three days later, referring to me as a "cybercop" and a "cyber-
sleuth," I had never intended to be one. In the winter of 1995
the only thing I was aspiring to be was a ski bum, and so far I
wasn't doing a very good job of it. During the best ski season in
memory in California, here I was stuck on a chilly morning in a
parking lot in a suburb of Raleigh, North Carolina, a long way
from anything resembling a ski trail.

   I was holding an antenna that looked a little like a ray gun in
one hand, and in my lap I was cradling a device that resembled
an oversized electronic daytimer. It was emitting a soft whistling
tone, much like that a modem makes when it establishes a con-

   The sound from the device had become persistent-proof I'd
cornered my quarry, an elusive computer outlaw who'd managed
to stay one step ahead of the FBI and at least three other law
enforcement agencies for more than two years through a combi-
nation of con-artistry and sheer luck.

   Along the way I'd been one of the victims. In December he
2                            PROLOGUE
and possibly some of his cronies had electronically broken into
my computers and stolen software I'd written, which if abused,
could wreak havoc on the Internet community.

   Now I was in a position to even the score. But none of us in
the family van were police. The driver was a beefy engineer for a
cellular telephone company and in the back sat a New York Times
reporter who had tagged along following my odyssey. Ten min-
utes earlier our van had slowly circled a nondescript apartment
complex as I swept my antenna back and forth, intently watch-
ing a digital signal-strength display for signs I was getting closer
to the source of the cellular telephone call. I was determined to
end my pursuit, but now, through a fog of fatigue brought on by
an almost sleepless week of pursuing a vaporous trail of digital
footprints through the web of computers that make up the
Internet, I could feel the paranoia that can well up inside you
when you push yourself too hard and too far.

   Outside it was dead quiet. There were no cars or people on the
street, and I felt conspicuous as our vehicle slid quietly by the
apartment complex under the yellow glare of sodium vapor
lights. Where was he? Was he watching us? Was he about to flee?
As I watched the meter readout it suddenly fell off He was
behind us. Was he on the other side of the building? The driver
turned the corner, and we saw empty fields stretching away into
the darkened countryside. Our maps which we had spread out
before us in the van showed a state park.

   ''A perfect escape route," the reporter murmured from the back

   We turned another corner, and the van swung back toward the
front of the complex. The antenna swept back and forth and inside
the darkened cab I watched as numbers on the display flickered
upward again. Toward the front of the building our van slowed,
and we crept through a parking lot full of empty cars. As we neared
the corner of the apartments we stopped briefly. Using radio detec-
tion finding gear is a little like playing "pin the tail on the donkey."
You get little cues, but still you feel like you're flying blind, floun-
dering around in the dark. Now, however, from the way the meter
jumped, I could tell we were almost on top of our target.
Somewhere, within thirty meters of us, someone was crouching
over a computer with an open connection to the Internet. It was


     impossible to decipher the meaning of the monotonous hiss that
     was proof he was still at his keyboard. Where was he right now?
       The three of us craned our necks and peered into a cul-de-sac.
     From inside the complex came a light from a second-story win-
     dow. How would a fugitive react if he peeked outside in the mid-
     dle of the night and saw a van with an antenna inside idling in
     his driveway? It was obvious he'd flee, or maybe worse. I had no
     idea what his state of mind was. Was he alone? There was no rea-
     son to think our cyber-criminal was armed, but it was late, and a
     cold feeling of doubt was growing in the pit of my stomach.
       "If I was him I'd be facing the window," the reporter suggested.
       He was right-we might have blown everything. Weeks of
     painstaking, cross-country detective work would vanish, leaving
     us empty-handed and chagrined. We decided caution was in
     order. The van started moving again and rolled around the cor-
     ner of the building.
Is it possible to drive the 310 kilometers west from Echo
Summit on the crest of the Sierra Nevada range to San Francisco
International Airport in under two hours?
  On the day before Christmas 1994 I tried it-in a snowstorm.
  I thought I had a good reason. I was eager to see a friend
whom I hadn't seen for more than two months and I had been
feeling unsettled about where our relationship would be when
she returned from her travels. We had been close friends for
three years and during the last six months it had become clear
that we were more than friends; we were in love. We had both
agreed that during our time apart we would think about where
we wanted to go with our relationship. Now I was in a hurry,
because I was full of anticipation, but at the same time, I was
also nervous and uncertain. What I didn't realize was that my
headlong race from one side of California to the other was the
opening act in an unusual adventure that was about to change
my life forever.

  The previous afternoon, Julia Menapace had left a message on
my home answering machine in San Diego: she was at the airport
in Bangkok and would be arriving in San Francisco at 1:40 the fol-
lowing afternoon after a fourteen-hour flight. Would I meet her?

   Of course I would. I had been thinking a lot about Julia; the
message suggested she'd been thinking about me too.

  A tall, graceful woman who is strong and wiry, and who often
wears her hair drawn back in a braid, Julia had been a program-
mer at Apple Computer and other high-tech companies in
Silicon Valleyfor the better part of a decade. With an intense gaze
and blue-gray eyes, Julia was often introspective but also quick to
laugh. She was a talented yoga teacher and had an ethereal qual-
ity that I found completely captivating. Recently she'd been
working as an independent programmer, brought in by high
technology companies to work on specific software development
   Although she knew the inner workings of the Macintosh com-
puter well, she never become as obsessed with computing as
many of the men she worked with. She had never been com-
pletely sucked into Silicon Valley's round-the-clock hacker cul-
ture-she liked to do too many other things in life, away from
the computer world where time is divided into nanoseconds.
During the years we'd known each other, we'd gone on countless
trips, exploring the backcountry: mountains, hot springs, beach-
es. We both shared a love of the wilderness, whatever the season.
   Julia had a particular passion for the mountain world above six
thousand meters and in the fall of 1994 she took off for the
Himalaya, but before she left to climb and trek in Nepal we had
a great adventure exploring the Southwest together. We hiked
Bryce Canyon and Zion National Parks and wandered among the
Anasazi ruins in Chaco Canyon. It was during trips like that one
that I'd corne to see Julia as the wonderful person she is, and we'd
fallen in love. I knew she wanted to be in a committed relation-
ship, but I had told her I needed to think about whether I was
ready for a serious partnership. We hadn't spoken since right after
she arrived in Katmandu, but after a couple months of contem-
plation I'd decided that I wanted to be with her and I thought I
was able to uphold my end of a partnership.
   However, I had no idea ifher thoughts were tracking mine, and
our relationship wasn't simple. Things remained ambiguous
because she was also trying to end a seven-year-old relationship
that had been drawing to a painful close for a long time. The man
she had lived with had once been a friend of mine-a Silicon
Valley hacker and a privacy activist who was well known for his
commitment to making sure personal privacy wasn't lost in the
emerging digital age. It had been a painful time before Julia left
the country, but it was clear to me that their relationship hadn't
                         JULIA'S RETURN                            9
been working and it was a question of when, and not if, it would
   But I didn't know what was going to happen next. I'd missed
Julia and was eager to see her. It was important for me to arrive
at the airport on time-getting there, however, meant coming
from the eastern side of the Sierra Nevada close to the Nevada
border. Just a day earlier, I'd moved in to an A-frame cabin out-
side Truckee, California, a couple hundred meters from the
Tahoe-Donner Ski Resort, in the midst of a cross-country skiing
mecca, with Emily Sklar, a ski instructor who had been a good
friend for several years.
   In San Diego, where I work most of the year, I in-line skate for
exercise, but as much as I like to skate, I like cross-country skiing
even better. During the last three years I'd learned a cross-coun-
try skiing technique called skating that looks much like in-line
skating and offers more speed than the traditional striding tech-
nique you see most skiers using. Instead of skiing in two narrow
tracks, skaters glide forward, placing each ski diagonally to the
trail. I also like to race and the previous winter I'd begun to take
racing seriously again, and had raced in several biathlons, a com-
bination of skiing and riflery that combines strength, speed, and
   Of course snow isn't one of San Diego's strong points. The pre-
vious winter, the ticket agents and flight attendants on Reno Air
got to know me well. Once I even packed an ice ax in my carry-
on luggage and sent it through the X-ray machine. No one
blinked. In that single ski season I logged more than thirty thou-
sand kilometers between Southern and Northern California. My
plan this year had been to spend the winter skiing, volunteering
for the Nordic ski patrol, serving as a parttirne ski instructor and,
when time permitted, taking on intriguing research problems.
   The kind of work I often do, computational science and com-
puter security research, can be done from just about anywhere.
And because the previous winter I'd found myself flying up from
San Diego nearly every weekend, this year I'd decided simply to
set up headquarters in the mountains for four months. I planned
on bringing along a couple of Unix workstations and connecting
my own computer network to the outside world with a high-
speed digital telephone line.
   Usually I spend most of each year wearing several hats. Until
I0                          BREAK-IN
the winter of 1995 I was both a senior fellow at the federally
funded San Diego Supercomputer Center-a facility on the
University of California at San Diego campus-and a research
scientist in the university's physics department. The Center pro-
vides me with an office and access to some of the world's fastest
supercomputers. My work has always involved doing research in
an area that has fundamentally transformed science in the last
two decades-the physics of computing. Computation has
emerged as a third way of conducting science, taking its place
alongside traditional theoretical and experimental methods.
   Where it was once necessary to prove scientific theories by
doing real-world experiments, computers have become fast
enough that it's now possible to create accurate simulations of
real-world events. Computational physicists attempt to solve sci-
entific problems through simulation. Ever more powerful com-
puters make it possible to realistically simulate everything from
the flow of air across the top of an airplane wing to the basic
structure of matter in the hunt for the top quark.
   Computational physics is also about the physics of computing
itself, discovering how electrons can be marshaled to manipulate
ever vaster stores of information in ever faster ways, and about
designing specialized machines that surpass the performance of
today's best supercomputers. Like many people in my field who
were trained first as physicists, I have begun spending more and
more of my time in recent years on real-world computing prob-
lems like computer security. In one sense it's a grand tradition
among both physicists and computer hackers. Nobel Laureate
Richard Feynman was notorious for his safecracking escapades at
Los Alamos during the Manhattan project days. And Robert
Morris, one of the inventors of the Unix operating system and
later the chief scientist of the National Security Agency was a pio-
 neer in understanding how to break in to computers as well as
protect them.
   I've always found it a compelling intellectual challenge-find-
 ing the chinks in the armor of a computer or computer network
 that, unprotected, might enable a digital thief to loot a bank's
 electronic funds or permit foreign spies to slip into the Pentagon's
 computers. It's a world that you can't approach just on an acade-
 mic or a theoretical level. You have to get your hands dirty. The
                         JULIA'S RETURN                           II
only way you can know for sure the digital locks are strong
enough is if you know how to take them apart and completely
understand them. My research into different paradigms for com-
putation has provided new tools to evaluate strengths and weak-
nesses in computer networks.
   Until I decided to move my base of operations to the moun-
tains for the winter I had been doing more and more research on
computer security at the San Diego Supercomputer Center, or
SDSC, where the tone was set by the director, Sid Karin, a tall,
thin, bearded, and unflappable former nuclear power engineer in
his mid-fifties. Like many people who have taken circuitous
routes into computing, Sid was working at General Atomics, a
Southern California-based nuclear power plant contractor,
when he decided he could do a better job developing the com-
plex simulations necessary for power-plant design than the pro-
grammers who were assigned to the project. One move led to
another, and today he is running the Center, a four-story build-
ing, housing a Cray C90 and Intel Paragon supercomputer, with
the mission of pushing the frontiers of high-powered computing
and pure science.
   The Center itself, an antiseptic white four-story building set on
a hillside on the university campus, is no model of architectural
splendor, and we refer to it as "the box the building came in." But
it's a reasonable place to do research, and it attracts a lot of peo-
ple who don't like regular hours or bureaucratic routines. Sid
barely blinked on the evening I skated into his office.
   Which is not to say I haven't managed to rub some people at
the Center the wrong way. I had an early run-in, for example,
with the deputy director of operations, Dan D. Drobnis, whom
I and others refer to behind his back as "D3."
    One day in 1992, D3 discovered me skating in the machine
room, the sprawling glass-enclosed space where the Center's main
hardware is housed. He went totally nonlinear, insisting I would
crash into one of his multimillion-dollar computers and swearing
I would never set foot in the Center again if I came anywhere
near the building in my skates.
    It seemed like an extreme and unreasonable attitude. Since I was
constantly crossing the room, moving between the front door and
a specialized graphics workstation some thirty meters away, I
12                          BREAK-IN
thought skating made perfect sense. But I can be a pragmatist in
some matters, and since that incident, I haven't exactly avoided
D3, but I haven't skated into his office, either.
   The worst excesses of the bureaucracy aside, life at the
Supercomputer Center has for the most part been a reasonable
compromise. But in December 1994, I had vowed that things
would be different. Truckee, where my ski cabin is located, is
twenty kilometers from Lake Tahoe, and the country around it
has the advantage of being both high enough to catch the most
snow and reasonably dose to Silicon Valley, where many of my
computer-security sponsors are located. But to get there from
the lake region you typically have to cross the notorious Donner
Pass, where the wagon train of the Donner Party became
snowed-in in October of 1846. It was completely illogical for
them to attempt a crossing that late in the season. Trapped by
heavy snows and facing starvation, some of the pioneers resort-
ed to cannibalism, and only about half of the original eighty-
seven members survived.
   It's a story that every California schoolchild is taught to illus-
trate the hardships endured by their gritty forebears. These days,
though, most of the skiers who flock to the area each winter tend
to pay little heed to the elements. I know a Silicon Valley software
engineer whose favorite T-shirt reads "Donner Pass, Calif. Who's
for Lunch?" But on that day before Christmas 1994, I developed
a new respect for the Donner Pass.
   I probably should have set out the night before, and had in fact
briefly considered leaving then and spending the night in the city.
But the weather looked like it was about to become snowy and
miserable, and I was tired from a day of skiing, so I drove back to
the cabin and went to sleep.
   It was about 8:30 on the morning of December 24 when I
swung my rented Ford Probe out of the slushy driveway of my ski
cabin. A light snow was still falling, but I didn't plan to get out
of the car until I was down out of the mountains, so I was dressed
for California winter: T-shirt and Patagonia shorts, Oakley sun-
glasses, and Teva sandals. I was giving myself plenty of time for a
leisurely trip through Donner Pass on Interstate 80, down the
foothills, across the Central Valley, south on the freeway through
Berkeley, over the Bay Bridge, and then south through San
                        JULIA'S RETURN                          I3
Francisco to the airport on the west rim of the Bay. I figured I
would be there by 11:30-noon, if I stopped for a strawberry
milkshake at Ikeda's in Auburn.
   Soon after I left I dialed Caltrans for the recorded highway
report on my cell phone and got the bad news: there was "chain
control" on Interstate 80 through the mountains. That meant
that it was snowing much more heavily up ahead and that the
California Highway Patrol would be stopping cars to see if they
had chains, and if they didn't, be turning them back. My rent-a-
Probe, of course, was chainless.
   The report did say that Highway 50, the road that stretches
from Sacramento to the south end of Lake Tahoe, was still open.
I turned around and headed in the other direction, past the
Squaw Valley ski resort and along the California side of the lake.
But any hopes that I might skirt the storm, and scoot past chain
control on Highway 50 evaporated when I reached South Lake
Tahoe 90 minutes later. Ahead of me stretched a long line of cars
stuck at the CHP chain control station.
   I was beginning to understand how ill-equipped the Donners
must have felt when they realized spring wouldn't come nearly
soon enough. I whipped the Probe around and raced into town.
Fifty dollars and an hour later I was in the chain-control inspec-
tion line with everyone else waiting to creep over Echo Summit
on Highway 50.
   It was almost 11:30 before I really got moving. Ford Motor
engineers take note: your basic Probe will do 130 kilometers per
hour with chains on, although it doesn't sound pretty.
   I have a radar detector, which is great when you're flying along
open stretches in Nevada. But in California a detector isn't very
useful, for the CHP has found an effective low-tech way to beat
speeders with these gadgets. Instead of using radar, they simply
race down a highway entry ramp in their black-and-white inter-
ceptors, pace the unwary speeder long enough to clock him, and
then smugly pull their quarry over.
   On that day, I was either extremely lucky, or all the CHP cars
were too busy enforcing chain control to worry about speeders.
   I called on the way down to check on the arrival time of Julia's
United shuttle flight from Los Angeles. It looked like I might be
there late and so I asked the airline to deliver a message to her.
I4                         BREAK-IN
The message missed her in Los Angeles and so 1 called again and
asked United to deliver a message to her on her flight, and they
said they would take care of it.
  The trip was more than 300 kilometers of California freeway
driving, and 1 figure 1 averaged 155 kilometers per hour-a bit
slower with the chains on-for the first 130 kilometers of
Highway 50, certainly much faster after 1 stopped and took the
chains off.
   By 1:30, 1 had managed to arrive, park, and position myself
just inside the airport security checkpoint just as Julia, in her
lanky stride, came down the escalator in the United Airlines ter-
minal. By her expression 1 could tell she was surprised to see me.
   "1 take it you didn't get my message," 1 said.
   "What message?" she replied. But it didn't matter. We hugged.
Later she said 1 had looked a little harried.
Of all the questions raised by the first attack, one still puzzles
me: was it simply an extraordinary coincidence that the initial
raid was launched from Toad Hall?
   Toad Hall, an exquisitely renovated two-story Queen Anne
just north of San Francisco's Haight-Ashbury district and
Golden Gate Park, is owned by John Gilmore, a Unix hacker,
libertarian, and electronic privacy activist. John had also been
the fifth employee of Sun Microsystems in 1982, years before it
became a publicly held company and one of the world's leading
makers of workstations and networking systems. He left Sun
four years later but the millions he made from being one of the
first employees at one ofAmerica's most successful companies let
him purchase a beautiful home.
   The name he chose for the place originally comes, of course,
from the home of Mr. Toad in the Kenneth Grahame children's
book classic The Wind in the Willows. "Toad" also happened to
be the nickname of a woman John was living with when he
bought the house. Either way, the name was a suitable title,
because the fictional Mr. Toad was a wealthy free spirit, and so
was Mr. John Gilmore.
   With John and friends in residence, Toad Hall became a pro-
totype: one of the first digitally networked homes in San
Francisco, the city where new social trends always seem to be
accepted first. In the fifties it was the beat generation, in the six-
I6                           BREAK-IN
ties the hippies, in the seventies alternative sexuality, in the eight-
ies it was south of market skateboard punks. Now in the nineties,
cyber-communes seemed to be sprouting up all over the city.
   In this scheme a group of starving artists, or bicycle messen-
gers, or even financial-district hackers will get together and rent
a house or a flat or an apartment in order to pool their money to
share a fifty-six-kilobit-per-second line, leased from the phone
company for several hundred dollars a month, to connect to the
Internet. Or if the group is more solvent, they might scrape up
several thousand for specialized hardware and perhaps a thousand
dollars a month for an even faster T-1 connection.
   A T-1 line can give you a garden hose of computer data from
the Net, compared to the strawlike modems that most people
use to connect to on-line services like Compuserve, Prodigy,
and American Online. A T-l data line will transmit 1.5 million
bits of information a second. That's enough to download the
complete text of Moby-Dick in twelve seconds, or to watch a
full-screen movie in real time. (Before things get really interest-
ing, however, digital network speeds will need to increase by
roughly two orders of magnitude-the equivalent of a fire
hydrant-something unlikely to happen until after the turn of
the century.)
   I take the Net for granted as part of my work, but I can under-
stand why people who must pay their own way might want to co-
op. Still, the commune idea seems an odd one to me. If the
Internet is about building "virtual communities"-electronic col-
lections of people with no face-to-face contact-doesn't it seem
strange that they feel a need to live together, too? In any case,
John Gilmore was starting a trend, not following one, when he
moved into the Queen Anne in 1987. It had two flats, one for
himself and his girlfriend, the other initially for use by a friend he
eventually bought out. From the beginning, this was to be no
mere residential building; it was a place to live on-line. A coaxial
Ethernet networking cable soon snaked its way through the
entire house. There were also computer workstations occupying
various places, from bedroom desks to basement tables, for use by
the various residents, house guests, and drop-in visitors who lived
in and hung out in Toad Hall. Where other people might put a
coat rack, in the entrance hallway to his second-floor flat, John
had placed a Sun SPARCstation ELC.
                           TOAD HALL                            I7
   In keeping with Internet nomenclature, Toad Hall acquired the
Internet domain name toad.com, whose gateway to the rest of
the world was a Sun SPARCstation computer in the building's
basement. This digital domain was run by John and an eclectic
band of programmers and hardware gurus, who together had a
diverse political outlook, and while privacy was a priority, com-
puter security at Toad was often pretty loose.
   John's Toad Hall experiment eventually spawned an early
Internet cooperative called The Little Garden, named after a
Chinese restaurant in Palo Alto where the first organizing meet-
ing took place. Started by a well-known San Francisco computer
hacker named Tom Jennings, The Little Garden was one of the
first low-cost ways of getting directly hooked up to the Internet.
Unlike today's live-in cyber communes, though, The Little
Garden didn't require being physically located in Toad Hall to
enjoy its electronic benefits. A member bought two modems and
put one at his home and one in the basement of Toad Hall. This
second modem was connected through a router to Toad Hall's
network link to the Internet and as a result members were per-
manently on the Net.
   The setup was economical, because Pacific Bell offered
un metered residential phone service. So it was possible from a
business telephone to leave your data line connected around the
clock for a single monthly fee that members chipped in to The
Little Garden. If the line was dropped, your modem at the Little
Garden would call you back at no charge. Toad Hall eventually
had more than a dozen phone lines running into it, and the Pac
Bell installers probably wondered what kind of boiler-room
racket John and his gang were running in there.

For the past five years, Toad Hall had been Julia's home-for
John Gilmore was the "other man," with whom her relationship
had been souring even before she and I had met. During the
Christmas holidays John was away visiting his relatives in Florida,
and so Julia and I had Toad Hall to ourselves when we arrived
around 4 P.M. on the afternoon of her flight from Nepal.
  John, now forty, was someone I'd known from hacker circles,
and even as a friend, for a number of years. Several years earlier
he had helped found a second company based on some of the
I8                         BREAK-IN
principles of an organization called the Free Software
Foundation. The idea behind the company, known as Cygnus
Support, was not to sell software directly but, instead, to give it
away and then sell the support and maintenance that corpora-
tions would require to make full use of the programs such as
computer languages and security tools that Cygnus developed.
It's a powerful idea, and the company was thriving, even in a
world dominated by Microsoft.
   A thin man with shoulder-length blond hair and a beard, who
sometimes wore the flowing flower-child clothes that had been
in fashion in the Haight-Ashbury during the sixties, John had
thrown himself into the new start-up with a passion that con-
sumed most of his waking hours. Initially he hadn't minded that
Julia and I spent time backpacking together while he worked
long hours on his new start-up, because hiking didn't interest
him. But once Julia and I had become more intimately involved,
things grew chilly between him and me.
   Julia and I sent out for dinner from an Italian place called
Bambino's. When it came, we undressed and sank into the indoor
hot tub, eating while we soaked.
   The upstairs bathroom in Toad Hall is an unusual room. It is
faced with a white and pink marble floor and wainscoting sur-
rounding a dark green jacuzzi tub and other fixtures. A large
asparagus fern sits on the window ledge, centered above the cas-
cading waterfall of the tub's larger faucet. The fronds of the fern
tumble down toward the water. Julia had put on a cassette tape
of Karma Moffet playing Himalayan intruments, and then lit
candles; the only other light came from four overhead spotlights
that dimly illuminated each corner of the tub.
    "This is just amazing," Julia murmured through the steamy air.
She said she had fantasized continually about a long soak in hot
water while trekking in the frigid Himalayas, where water is car-
ried by hand from its source and becomes hot only when heated
over flames, and where there is never enough to sit in. And at
high altitude in the Solu Khumbu region of Nepal, the only heat
had come from the sun, the small cooking fire, and the occasional
woodstove fueled by wood scraps or dung.
   While we ate Julia told me stories of her adventures. In the
kitchen of a lodge where she stayed she met and befriended a
                           TOAD HALL                            I9
Sherpa guide named Tshering and a mountain guide from Seattle
named Rachel DeSilva, who had led a group of 12 women to
climb a 6,000-meter trekking peak in the region named Mara.
Mterward they had invited her to climb another mountain
named Lobuche, which lay to the north toward Everest. She had
made it to just below the summit.
   I sat entranced. "I wish I had been there too," was all I could
find to say.
   Julia had spent her birthday at the Tengboche monastery to cel-
ebrate the Mani Rimdu festival. She showed me a red string neck-
lace that she had received when a Tibetan Lama had blessed her
on her thirty-fifth birthday.
   "Near noon that same day, I heard the sound of long horns,
cymbals, and drums," she recalled. "Then an avalanche poured in
slow motion down the south face of Ama Dablam."
   Later in the trip she had stopped at one point to watch a sun-
set over Everest through the gathering mist, and she said it was so
stark and beautiful that she cried. "I thought of you," she told
me, "and wished you were there to share it with me."
   As we soaked, I told her about what had happened to me
while she was gone. When Julia left I had been waiting for a
$500,000 per year research grant from the National Security
Agency, the nation's electronic intelligence-gathering organiza-
tion. The NSA has two missions: one, its foreign spying mission
and the other its responsibility for the security of all the gov-
ernment's computers and communications. In the fall an infor-
mation security division in the agency had told me they would
fund a project permitting me to assemble a team of experts to
do research in new areas of computer security. I was ready to go
and I had commitments from people to start work, but the
agency had dragged its feet for months. Finally I had gotten
tired of being jerked around, and two of my researchers had
been forced to take other jobs.
    "I thought everything would be ironed out and I'd come back
to find you happily at work with your team," she said.
    "No it wasn't," I answered. "They're amazingly inept, just like
 any government bureaucracy. "
   We talked for a while about the NSA and how so many people
 in the civil liberties community fear them as Big Brother as well
20                          BREAK-IN
as anyone associated with them, arguing that they become cor-
rupted by association. But that had never seemed accurate to me.
Everything I'd seen indicated they were a largely incompetent
organization tied up in endless regulations that could do little
good or evil. And people are quite capable of making up their
own minds.
   "I don't want to deal with them," I said
   ''I'm sorry it didn't work out, Tsutornu," she said quietly.
   We soaked for a while, both of us lost in thought. Finally I
changed the subject.
   "I want to tell you something I've been thinking about," I
said. "I've thought about a lot of things while you were away. I'd
really like to try having a committed relationship with you, if
you're willing to."
   Julia smiled. She didn't say anything, but she reached over and
held me closely.
   It seemed like we would now be able to share a lot of time
together. I told her I'd taken a leave of absence from the univer-
sity and now I was looking forward to skiing and getting away. I
was finally pursuing my long-held plan to spend a winter in the
mountains, spending the mornings and late afternoons skiing
and the mid-days and evenings thinking and working on my
research projects.
   "Why don't you come with me and live in the mountains?" I
suggested. "You can come ski and it will be good to be outside."
   We woke at about 1 P.M. the next day and Julia-who grew up
on the East Coast and is still learning to deal with mild California
winters-told me that she had seen the first morning light before
she fell asleep and thought to herself, It's Christmas and there is no
sign of it here. She was still jet-lagged and also feeling what she
feared was flu coming on. We decided to spend the day inside,
catching up on talk and sleep. It was chilly out, so Julia turned up
Toad Hall's central heat, still eager to soak up the warmth of civ-
ilization after two months in the Himalaya.
   A bit later, while she rested, I was walking around the house,
and several times went past the Sun SPARCstation in the hallway.
It was a reminder that I probably had new electronic mail, but I
didn't feel like checking it.
   At just about that moment, however, ominous bits of data were
                           TOAD HALL                            2I
flowing through the Ethernet cable that wound through Toad's
rooms and hallways. From somewhere, perhaps thousands of
kilometers away, an electronic intruder had taken control of
toad.com by remotely commandeering the SPARCstation in the
basement. And while the two of us spent the day together two
floors above, the electronic hijacker was using toad.com as a stag-
ing base to launch an attack on the computers in my own beach
house some 800 kilometers south.
   I didn't realize it that afternoon, but the intruder had made
himself "root" of toad.com. The root account is an all-powerful
computer system administrator, an entity who can control every
operation of a Unix machine. It is usually reserved for a comput-
er's caretaker or administrator. On a Unix computer like the
SPARCstation in the basement of Toad Hall, being root is like
being God. Once he has become root, a computer user can cre-
ate and delete accounts and files, read any other user's mail or
documents, monitor someone else's every keystroke, or tamper
with a computer's software to leave behind programs that create
secret backdoors for easier entry next time.
   Whoever it was that invaded the system had a reasonable
degree of computer network sophistication-or at least an aware-
ness of the lax security of toad.com. Whoever it was also had
obviously picked out my computers in San Diego as a specific
target, whether out of a personal vendetta or because they
assumed I had valuable files.
   As one of a small group of high-level computer security
researchers in the country, I have machines that store sensitive
information, like reports on little-known bugs, loopholes, and
system vulnerabilities that have been discovered on various
types of widely used hardware and software, and I also have a
range of computer security tools. However, I had taken many
security precautions, and material that I considered extremely
valuable wasn't accessible. Still, some of the information and
tools were within reach of the determined intruder, and in the
wrong hands they could be used for breaking into other civilian
or government computer systems, or sold on the corporate espi-
onage market.
   That evening we ordered dinner out again-Indian food this
time. While we waited for the food to arrive, Julia began to
22                          BREAK-IN
unpack from her trip and I spent some time setting up a new
portable computer I had picked up from a friend's house the day
before after leaving the airport. Made by RDI, a San Diego-area
company I consult far, it was a compact Unix workstation and I
had volunteered to test the new model for them. I briefly thought
about connecting it to the Toad Hall network, but I didn't. I had
no idea someone was spending Christmas Day committing a
felony over the Internet.

Julia woke the next morning still feeling sick, so instead of going
for a day hike in the Headlands across the Bay in Marin County,
as we had planned, we spent another quiet day at Toad Hall. It
was cold and grayout, and the only time we left the house was
briefly at midday, when we wandered over to Haight Street to
have lunch at Cha Cha Cha, a tapas place that attracts an eclec-
tic crowd, ranging from those who live in the Haight to white
collar financial district types and people ofvarious ethnicities and
hues from all over the city. John was arriving home that evening,
and there were clearly matters that he and Julia needed to talk
about. I had things to do in the South Bay area and if everything
worked out well, in a few days Julia would come skiing with me.
   ''I'll see you soon, I love you," she said as I went to the door.
   "Take care of yourself," I said, and we embraced.
   Shortly after 8 P.M., I climbed into the rented Probe and began
the 50-kilometer drive south to Silicon Valley, where I had
arranged to visit a friend named Mark Lottar. A young hardware
hacker and Internet whiz, Mark, who was thirty-one years old,
was a friend with whom I'd spent a lot of time exploring cellular
telephone technology. Mark is on the short side with brown hair
falling carelessly across his forehead, but he has an adventurous
hobby: hopping freight trains for an occasional hobo's sojourn
around the West. Most of the time, though, he is attending to
Network Wizards, the small company that he runs from his
home in Menlo Park, which makes and sells a variety of useful
computing tools ranging from computer temperature sensors to
cellular telephone network diagnostic and surveillance tools that
are popular with cellular companies and law enforcement offi-
cials. Together we had taken apart the software that sits at the
                           TOAD HALL                             23
heart of Oki's cellular telephone. Mark had originally spotted it
as a well-designed piece of technology, and I'd read his review of
the Oki 900 and bought one as well. Once we knew how the soft-
ware worked we figured out how to control it with a personal
computer. For just a little more than one hundred dollars his
hardware and software allows an Oki and an inexpensive personal
computer to rival bulky commercial diagnostic products that cost
many thousands of dollars.
   Most people know Mark for his biannual survey of computers
directly connected to the Internet, the electronic equivalent of a
Commerce Department census. He has written software that sys-
tematically "walks" through the Internet, querying each major
computer domain. As with the human census, many computers
choose not to answer, but his numbers are the best basis for an
educated guess about how big the Internet is and how quickly it's
growing. His most recent survey in mid 1995 counted 6.6 mil-
lion computers with a Internet connection. Of course that num-
ber doesn't indicate how many people actually are on the
Internet, because one computer with a direct connection to the
Internet may be the network gateway for tens, hundreds, or even
thousands of users and their own individual computers. Still, it is
from Mark's survey that most estimates, from the conservative to
the giddy, are derived.
   As I drove I was feeling a bit harried-I was late for meeting
Mark and some friends for dinner and I was still thinking about
Julia. US 101 took me south out of San Francisco passing
Candlestick Park, the airport, and the rim of the Bay with its
light-industrial sprawl, which forms the northern reaches of the
Valley. The road was wet from a recent, chilly rain, a good sign. It
meant more snow up in the mountains. My plan was to pick up
Julia in a day or two, then return to the Sierra and a winter that
was looking as if it might offer the best skiing in years.
   It was shortly after eight o'clock and I was nearly to the
Highway 92 overpass, the unofficial northern border of Silicon
Valley, when my ringing cell phone broke my train of thought.
   "Tsutomu, this is Andrew." He needn't have identified himself,
for his voice, with its residual trace of Tennessee vowels, IS
 instantly recognizable.
   "Have you got a minute? Can you get to a land line?"
24                         BREAK-IN
   "Not conveniently," I answered. He was spending Christmas
vacation at his parents' house in Tennessee. Andrew Gross was a
twenty-seven-year-old electrical engineering graduate student at
the University of California at San Diego who worked with me
on networking and security problems at SDSC (the Center). He
had great promise as a computer security researcher, and I'd
become something of a mentor to him. As part of his appren-
ticeship he often watched over my network when I was away. As
we talked I got the distinct impression he was really uneasy talk-
ing to me while I was using a cellular phone and at the same time
had something serious he wanted to tell me. I pressured him to
give me a basic idea in some way that wouldn't give anything
   "Tell me generally what's up," I said. In my harassed mood I
wasn't interested in dealing with new problems. He paused. He
was obviously mulling what he could say that wouldn't prick up
the ears of the few dozen bored or nosy people who were prob-
ably at that moment using radio scanners to monitor the local
cell phone airwaves the way some people listen to police radio
   "Well," he finally said. "Your log file summaries have gotten
   What he was telling me was that somebody had broken into
my computers. I had a queasy feeling, a little bit the way you feel
after you realize someone has picked your pocket. In my mind I
quickly ran through the implications though my immediate reac-
tion wasn't panic but irritation at one more distraction. We talked
for a while and gradually I realized that what he had discovered
was not an accounting error. Something was seriously wrong and
needed to be dealt with.
   My network is set up to maintain an audit file of all the con-
nections made from the outside world-a complete record of
who connects and when. Four times a day, a summary of this
information is routinely mailed to a remote computer that
Andrew monitored. Normally, the summary should get longer
between each mailing. If it got unexpectedly shorter, the logical
conclusion to draw is that someone had tried to erase the file.
   "Oh," I said, "shit," and thought for a moment about the best
course of action. "Why don't you dial in and see if you notice
                            TOAD HALL                              25
anything?" I suggested. "I'll go somewhere and see if I can learn
anything as well. I'll call you in a while."
   I keep a couple of modems connected to my computers that are
convenient for dialing directly into my network. It hadn't
occurred to Andrew to connect to our machines this way, but we
both knew that if we turned off our direct connection to the
Internet, then no one would be able to easily break in again over
the network, and the data on my computers would be more
likely to stay the way it was when Andrew discovered the short-
ened log files. He volunteered to mail the summaries of the
shortened log files to the wireless electronic mail terminal I usu-
ally carry with me.
   «Be careful," was the last thing I said before hanging up. «Make
sure you preserve the evidence."

Computer security is full of tradeoffs. The art form is coming up
with a set of tradeoffs you can live with. It is possible to have
absolute computer security: you simply unplug your computer
and lock it in a vault, and even the best thief won't be able to steal
your data. But this ultimate security solution also means that you
won't be able to use your computer. Like everyone else, I have to
make trade-offs in the security of my machines and take some
attendant risks.
  Although the Internet today is, as many writers have com-
mented, like the Wild West, with a lot of real outlaws roaming
through it, it wasn't always so. When I was in school at Caltech,
and later when I worked as a physics researcher at Los Alamos,
the world had not yet awakened to the Net. The culture still
showed its roots in the ARPAnet, the Pentagon-funded academ-
ic predecessor to the Internet that was established in 1969, and
was like a small community where everyone knew everyone else.
You greeted your neighbor at the general store and you left your
door unlocked.
   Today, with millions of people clamoring to connect to the
Internet, the rules have changed. The world is plunging ahead,
putting every conceivable form of business and private commu-
nication in electronic form and shuttling them back and forth
over networks that were originally designed for sharing informa-
26                          BREAK-IN
tion, not protecting it. There are, as a result, many tempting tar-
gets for bandits and information highwaymen.
   The vast potential for stealth is one of the most difficult aspects
of detecting a crime in cyberspace. In the physical world, if a thief
breaks into a bank vault, it's obvious there's been a burglary
because the money is gone. In cyberspace, a vault can be stripped
without any sign, at first glance, anyway, that a theft has even
occurred, for what is stolen is not the original piece of software
or data, but a copy that the thief makes. Even commercial soft-
ware programs worth millions of dollars can be copied in an
instant, leaving the original intact. It's all just bits.
   There is a school of thought in the Net community that argues
that, because a software program is infinitely copyable, then con-
ventional notions of property rights have little relevance.
Software should be free, they say, and freely disseminated, and
there should be no such thing as software intellectual property
rights. A chief proponent of this philosophy is Richard Stallman,
who has helped found groups like the Free Software Foundation
and the League for Programming Freedom.
   I believe that it should be up to the creator to decide whether
to give software away or to be compensated for the hard work of
writing it. And I certainly have no sympathy for those who per-
vert the free-software philosophy by reasoning that if it can be
freely copied it should be free to steal. There's no justification to'
illegally break into somebody's computer.
   In the last few years, as the Net has become more and more
commercial, computer vendors have begun selling security "solu-
tions"-protective hardware and software that is supposed to
make it impossible for vandals to break into your computer. The
problem with many security products, however, is that they are
stopgaps whose claims far outweigh their performance. Their
purpose is to make people feel better about security without real-
ly doing anything.
   One of the most common defenses is called a firewall, which
sits between the Internet and your computer and is designed to
allow only carefully screened bits to pass through to your own
network. Any data that is not recognized as friendly will be
blocked. The trouble with firewalls is that while they can be
highly effective filters, they can also make it cumbersome to
                            TOAD HALL                             27
use your computer on a network. In effect, they create a
Maginot Line instead of real security. A firewall provides a hard
shell but leaves the soft chewy center at the heart of your net-
work vulnerable.
   I refuse to be so paranoid that I can't function. My computers
are connected to the Net because it's a resource that not only lets
me share my work with a community of researchers, but also
makes an entire world of information-software, other comput-
ers, databases-accessible from my keyboard. Anything of any
sensitivity that I intend to send or receive over the Net is encrypt-
ed with software that makes it meaningless to anyone without the
key to the code. But my computers don't have any unusual elec-
tronic walls around them.
   Instead, I take other, less crippling precautions, like using
encryption, logging activity, and keeping log files, in some cases
with alarms. The real secret of computer security is to be aware,
to watch the systems carefully, something most people don't do.
   When intruders break into computers over the Internet, one of
the standard steps they take to elude detection is to erase traces of
their presence on the computer they've attacked. They will fre-
quently go into the automatic recordkeeping files, the log files,
that are generated by the system and edit out the record of their
   But this creates a situation that few computer intruders stop to
consider: when they erase the records of their activity, the length
of the file suddenly gets shorter. At SDSC and on my machines
at home, we have a simple electronic way of noticing this. When
Andrew connected to the Net to read his mail from his parents'
house in Tennessee, he had checked the summaries of our log
files, and realized we had uninvited guests.

It took me another twenty minutes to drive to Mark's two-story
townhouse. He lived in a building across the street from SRI
International, the research laboratory where the ARPAnet was
created a quarter-century ago.
   Inside Mark's house, the inventing continues. The place is full
of personal computers and workstations, all cabled together into
a local-area network. Like the cyber-communes in San Francisco,
28                         BREAK-IN
it has a T-l link to the outside world. Mark also has the perfect
hacker's totem in his living room: a 1950s-style Coke machine,
which provides a touch of classic industrial design. Most often
the machine is filled with bottled water, but sometimes it actual-
ly has Coke.
   Mark had been waiting for me. He planned for us to drive to
Palo Alto, a few kilometers away, to meet some of my friends for
dinner, but he saw that I was clearly distracted.
   "Sorry, " I said. "Something's up. I'll need a few more min-
utes." I explained briefly I'd had a break-in and wanted to assess
the damage.
   "I hope this won't take very long," Mark said. "I'm hungry."
   However, he was sympathetic; the previous fall he'd spent
weeks fighting off a determined electronic thief who was trying
to steal his cellular telephone software.
   I had a sense of urgency to get to things quickly, before the
data stored on my computer was lost or altered. Unlike a typi-
cal personal computer which until recently could run only one
program at a time, Unix computers generally run dozens of pro-
grams simultaneously, meaning if data were changed any trail
might be quickly erased. Normally I could have easily connect-
ed to my computers over Mark's network, but because Andrew
was about to shut off outside access via the Internet the only
option was using a modem to dial in directly to my computer.
I asked Mark if I could go upstairs and use his walk-in closet, a
tiny room in which he keeps clothes on one side and an IBM
PC and a low-speed 2400-baud modem on the other. Some
people hang on to out-of-style suits; Mark refuses to throw
away old-fashioned technology if it still has some wear left in it.
The security of the telephone was still an issue in the back of
my mind, but I decided the need to act quickly outweighed the
possible risk.
   I sat down in the cramped space and used the modem to con-
nect to my computers in San Diego. From Mark's PC I could
control the computers in my small network, both my comput-
ers at the SDSC and the ones I have at my house several miles
off campus. I browsed around for a while, looking through end-
lessly scrolling directories of files to see if anything was obvi-
 ously amiss. On the surface, everything appeared normal, so it
                            TOAD HALL                             29
was unlikely this was a mere prankster's break-in. As the altering
of our log files had indicated, somebody was trying to cover his
   I proceeded gingerly, as any detective would do, careful not to
damage any data that might later enable me to reconstruct how
the break-in took place. Even something as simple as reading a
file can forever obliterate an intruder's digital footprints. I could
see from the directories and the system administration logs that
Andrew was also connected to my network, making the same
kind of survey I was making-but with less care. He had been
tromping around, opening files to peek at them, and each time
destroying valuable evidence. It irked me, and I sent him a mes-
sage telling him brusquely not to disturb anything. But he had
spent almost an hour snooping around and information had
already been lost. Andrew's efforts, though, had led to one par-
ticularly significant discovery: some of the packet log files that
keep records of our network data traffic had been accessed recent-
ly, and then copied to an unknown location elsewhere on the
Net. This meant that whoever had broken into my machines now
could have any kind of information belonging to other users who
had used these machines, including their passwords. If I was the
locksmith, the thief now had lots of passkeys. I made a mental
note to go look through these data-traffic files later and do dam-
age control. Also, interesting but frustrating was Andrew's dis-
covery that the current network traffic log was invalid, and not
helpful to us.
   There was a range of information in front of us but none of it
seemed to form a complete picture. One hunch was to look at
copies of the logs from before they were erased and to see whose
records had vanished. From that we might be able to infer who
was trying to cover their tracks. We could see that around ten
0' clock the night before there had been a flurry of random
probes from a network site called csn.org, which was Colorado
SuperNet, an Internet service provider from which I had previ-
ously seen attempted break-ins. But none of the previous night's
attempts from csn.org appeared to have been successful. At
roughly the same time, we saw, there had been attempts to con-
 nect from two sites with prankish names: wiretap.spies.com and
suspects.com. If someone was trying to bait me, these were the
30                           BREAK-IN
type of taunts I would expect to see, although these clues weren't
putting us any closer to understanding what had actually hap-
pened. We also saw that one of my computers that handled pro-
gram communication over the network had been started for
some reason the night before. This was suspicious, but it might
not mean anything at all.
   I probed more deeply, very gently looking beneath the surface.
The file directories that a computer user sees are actually built
from other records a computer maintains at a much lower level.
By examining these minute details at the very lowest level of my
computer's file structure, I hoped to pick up some hint of the
changes that even the most clever interloper might not have
thought to erase.
   On Ariel, the computer at SDSC that serves as my gateway to
the Internet, I could see that the intruder had left some traces.
Much of the data wasn't even in English, but rather in the bina-
ry representations computers use to communicate internally, and
from it I could make out patterns of information still stored on
my computer's disk that revealed the ghost of a file that had been
created and then erased. Finding it was a little like examining a
piece of paper on a yellow legal pad: even though the top page
had been torn off, the impression of what was written on the
missing sheet can be discerned on the remaining page. The file
that had been there momentarily and then copied to some
remote point and erased had been named oki.tar.Z.
   It was a tiny clue but it was one that pointed in many possible
directions. What did it signify? Oki, of course, was the brand of
cellular telephone I'd worked on with Mark Lottor; it had been
the Oki source code-the programmer's original instructions-
that Mark's attackers had been after the previous fall. "Tar" is a
Unix utility program that archives and extracts files to and from a
single file called a tarfile, A tarfile is traditionally a collection of
files on a magnetic tape, but it can be any file. Someone might
have gathered software programs I'd written for controlling an
Oki cellular telephone, and then merged them into a single file
they called oki.tar. The "z" indicated they had probably squeezed
them down in size using another utility program "compress" so
that it would take less time to transfer them to some remote place.
   The fact that someone had bundled a bunch of files and called
them oki indicated a possible motive for why my computers had
                           TOAD HALL                             3I
been attacked: somebody was very interested in the inner work-
ings of cell phones. The ghostly shadow of oki.tar.Z also provid-
ed me with a set of directional pointers for determining which of
my files had been stolen. And because every file that had been
bundled together into oki.tar.Z had to be accessed in order to be
copied, and the access times had been duly noted by the com-
puter, I had a rather detailed accounting of my thief's visit.
   On the other side of the country Andrew only had one phone
line and so he disconnected and we talked by voice while I con-
tinued to examine my network from the computer in Mark's
   I told him the next step to take was to call the operations peo-
ple at SDSC and have them halt Ariel, the computer that con-
nects my network to the outside world and which is housed in
a wiring closet next to my office. To halt a computer is very dif-
ferent than shutting it down or rebooting it, both of which will
wipe out all the data that hasn't already been explicitly saved on
the hard disk. A halt command, by contrast, will freeze a com-
puter in its tracks, leaving every bit of information in exactly
the state it was when the machine was suspended. This step
would be crucial to the forensic analysis I now knew I was going
to have to perform, which also meant returning to San Diego.
Until we found exactly how my network had been attacked I
couldn't go back online. I was going to have to examine my sys-
tems with the software equivalent of a magnifying glass-or even
a microscope. And time wasn't on my side. What I needed to do
was in effect analyze footprints that had been left in sand. They
stand out, but only as long as they are not covered by others who
have traveled the same path.
   A little after nine my friends showed up, and finally, after nine-
 thirty Mark dragged me away from the computer closet. We all
 drove to The Good Earth, a health food chain restaurant in
 downtown Palo Alto. They needn't have waited, for all my pres-
 ence added to the party. I spent some of dinner talking to Andrew
 on my cell phone, trying to organize things so that we could meet
 as soon as possible in San Diego. He had already called Jay
 Dombrowski, communications manager at SDSC, and had con-
 vinced him that we'd had a serious breach. Dombrowski agreed
 that the Center would pick up the cost of the plane fare to fly
Andrew back to San Diego immediately.
32                          BREAK-IN
  There were few encouraging signs. By halting Ariel quickly
there was a chance we would be able to reconstruct some of what
happened, but log file information had been erased, and in our
brief examination we had not been able to turn up any obvious
backdoors, a telltale sign in many network break-ins.

Shortly before eleven Mark and I said good-bye to my friends and
we drove back to his townhouse in Menlo Park. I was still dis-
tracted, trying to come up with a plan to fly back to San Diego
and quickly assess the break-in. Back at Mark's, I connected to
my computers again and discovered that Andrew had spent more
time prowling through my network again after I had left for din-
ner. I could see he had done things that were likely to have erased
precious forensic data and I called him again, telling him that I
was upset that he had blundered around the network. I hung up
and realized that when I arrived in San Diego, I was going to
need hardware I didn't have to reconstruct the break-in. I asked
Mark to call his friend, Lile Elam, because I had cached some
disk drives and other equipment in her office at Sun
Microsystems several weeks earlier. I've been a consultant at Sun
for a number of years and Lile worked there as a technical sup-
port person, but the computer maker was closed down for the
Christmas holidays. I wanted to know if she was willing to meet
us at Sun for a midnight raid to pick up my gear.
   Lile was hesitant at first to go into the offices so late at night,
but I persuaded her to meet us in front of the building where she
worked at the company's Mountain View campus. I pointed out
that I knew everyone who mattered at Sun, and promised her
that if anyone questioned us, I'd take full responsibility for it.
When we arrived in front of Building 18 ten minutes later, Lile
was waiting outside.
   There was one problem. Parked right in front of the door was a
white Sun security pickup. Although the building was closed, it
meant a private security guard was probably wandering around
inside and might not take kindly to someone walking out with
handfuls of disk drives in the middle of the night. Worse, although
Lile and I had Sun badges, Mark didn't, and the security guard
might wonder what a non-Sun employee was doing with us.
                           TOAD HALL                             33
   We discussed waiting until he left, but none of us wanted to
spend the night in front of the building. We used our badges and
began walking down the hallway to Lile's office. Sure enough,
two-thirds of the way down the hallway we ran into the guard. As
it turned out, we were overly paranoid. He was happy with Lile
and me, but challenged Mark. We explained that Mark was a
friend of Lile's, though, and that seemed to satisfy him.
   Feeling a little as if we'd just crossed a border checkpoint, the
three of us went on down the hall to Lile's office where I gathered
up my disk drives and some interface cards in antistatic bags. We
could see the security truck through Lile's office window and a
couple of minutes later the guard left the building, got in the
truck, and pulled away. We hurriedly went back down the hall,
feeling conspicuous with our hands full of hardware. Outside we
walked in front of the video camera and got in our cars and left.
   When we got back to Mark's apartment around one-thirty, I
booked myself on the seven o'clock Reno Air flight out of San
Jose. I'd need to be up by six to get to the airport in time to drop
off the Probe at Budget. Mark wished me luck and headed
upstairs to bed, as I stretched out on the living room couch to
grab a few hours of sleep. I had one last thought before dropping
off: This year, skiing would have to wait.
Los Angeles has increasingly transformed itself into its own image
in the futuristic movie Blade Runner--a smoggy, anarchic tech-
   San Diego, where 1 live, isn't a pristine Southern California city
either, but it has a livable quality 1 never seem to find in Los
Angeles. It appears like an island every time 1 fly home on a com-
mercial jet dropping steeply over the financial district before
landing toward the ocean. Ringed by desert the city evokes a tan-
gible sense of the future drawn from the combination of the hard
angles of sterile twenty-first century architecture, lush palm trees,
bright green lawns, and ocean. There is no shortage of weird
architecture ranging from the abrupt modernist buildings on the
University of California campus, where 1 work, to the surreal
Mormon temple off 1-5 designed to evoke some mythical
European Renaissance church.
   When 1 left San Jose there had been long lines and mass chaos
and 1 was reminded of why 1 don't usually travel after three-day
weekends. On the first working day after Christmas the airport
was a seething mass of people all trying to get home. At about
eight-thirty on Monday morning, working on four hours of
sleep, 1 made my way out of the airport. 1 was loaded down with
thirty kilos of gear I'd picked up the night before, including the
prototype RDI PowerLite. 1 felt frustrated about being in San
Diego instead of on my way back to Lake Tahoe to ski. Andrew
was flying in a little later in the day so 1 walked out to the curb
                        DAMAGE CONTROL                            35
and took a thirty-dollar cab ride directly to the Center and my
   To describe it as an office is actually being kind. I have a small,
windowless room next to an even smaller wiring closet. It is
stuffed with several computer monitors, random hardware like
disk drives and other spare parts, and a Government security safe,
left over from my days as a federal laboratory employee, marked
travel so often I work in my office infrequently, but I use it as a
storage area. There are always lots of books and at least one tub
of unopened mail that my secretary has set down somewhere..
There are also several computer monitors connected by a bundle
of video cables to Ariel, the ancient Sun Microsystems worksta-
tion that is stuck away in the wiring closet. The closet is also
home to a few modems and several other computers including
one that functions as a routing computer-a traffic controller for
all the Internet data that comes my way.
   Several years earlier I had set up Ariel so that the background
on its video display constantly showed the most current satellite
weather image delivered over the Internet from the University of
Illinois at Champaign-Urbana.
   I've named my computers after the fallen angels in John
Milton's Paradise Lost. Unlike Byte magazine columnist Jerry
Pournelle, who seems to want to breathe life into his household
computers by calling them things like Ezekiel in his column every
month, I have no intention of anthropomorphizing the machines
I work with. To me computers are all basically "its." What I was
looking for instead were names that were obviously related but to
a casual observer not immediately obvious how. If it's done right
it becomes elegant. It also had to be a large set of names because
I was always needing more as new computers showed up.
Milton's fallen angels turned out to be a useful source of names
because I wanted something that offered lots of choices, and also
 something that would get by the network name censors.
   Before we had settled on the angels, Sid and I had had a chat
 about the naming problem, during which he told me, "I don't
want to censor you, but I don't want you to be offensive, either."
 Some people in the computer network world seem to believe I
 have an "attitude." Maybe I do, but this was just Sid's way of try-
36                          BREAK-IN
ing to persuade me not to pick unnecessary fights with the net-
work thought police.
  I tend to work on the fastest computers available at any given
moment, but I have something of a soft spot in my heart for
Ariel, which came with me from Los Alamos. It began life as a
Sun-B. Sun began shipping Sun 3s in 1985, which in terms of
computer workstation generations makes it an antique. New
microprocessor generations tend to come at about eighteen-
month intervals. Going back six generations of computer tech-
nology would be the equivalent of returning to the era of the
horse and buggy.
  Ariel has a curious history. Years ago I visited Sun with Brosl
Hasslacher, a Los Alamos physicist who has been my mentor over
the years, negotiating with Sun over problems we were having
with a much more expensive and powerful computer. A Sun exec-
utive went out on the loading dock and found Ariel for us as
something of a consolation prize. In the years that followed, Ariel
has become a vagabond, making its way back at one point to
Caltech, where it was used by a student who had worked with me
as an intern, before the machine finally wound up in my wiring
closet. Today I use it mostly for mail, for storing less-important
stuff, and as a jumping-off point to the Internet to give myself net
  As soon as I got into my office I dropped my bags. I glanced
at one of the monitors of the computer that had been frozen
the night before by a Center operator. On the display in the
console window, the one that reports system status informa-
tion, was an error message from the XNeWS interpreter, the
program that controls the display of information on the com-
puter's screen:

process (Ox480088, 'teal.csn.org NeWS client', runnable)
Error: Isyntaxerror
Command: '.psparse_token'

  I puzzled over the message briefly but could see no obvious vul-
nerability. I restarted Ariel long enough to peek at the statistics
logs kept in the modem that connected my computers at home
                       DAMAGE CONTROL                            37
to the Center. What I saw was evidence that there had not been
a lot of data taken from my home network. The modem log
showed it had been connected for five days-about the time I'd
been away-and that during that time about four megabytes of
data had flowed in each direction. That was just routine house-
keeping traffic and the fact that it was balanced in both directions
indicated that no one had siphoned files out of my home
machines. It was a relief-the principal target had been else-
where, probably somewhere among the computers next door in
the wiring closet.
  I called Sid Karim, the director, about the break-in. He was
generally sympathetic to my problem but was unwilling to give
me a blank check to solve the problem. He told me that if I was
describing the situation reasonably correctly then he was willing
to provide some expense money to help with damage control.
Translated politely, he was warning I'd better be right about my
suspicions. He also declined to pay my usual consulting fees,
arguing, "Tsutornu, you're basically on vacation."
  I decided that was about all I could hope for under the cir-
cumstances, and I went out to my Acura which I'd left in the
Center parking lot while I was traveling and drove home.
  I have computers both at SDSC and at home connected by a
high-speed modem line that always stays open. I'd decided to go
home first to examine things because that was where data and
programs I really care about are kept.
  My house is about a ten-minute drive north of the Center in
one of those newish communities of tract housing that dot the
Southern California landscape. The commute route takes me past
the Scripps Clinic and along what is described as San Diego's
Biotech Row. Just as Stanford University served as an incubator
for Silicon Valley, Scripps has nurtured a generation of biologists
turned entrepreneurs. My neighborhood was largely built in the
1970s and my house is actually a townhouse that fits snugly
against its neighbors. It's not my idea of the most desirable archi-
tectual style but it's close to campus and it gives me a sense of
being out of the city. I can see and smell the ocean and from my
upstairs bedroom I can hear waves crashing on the beach at night.
I also look out over Torrey Pines State Park where I go when I
need someplace to be by myself and think. The beach is out of
38                         BREAK-IN
both wireless computer data and cellular telephone range and
sometimes I will just take a pad of paper down there when I need
to concentrate.
  After parking in the garage I walked inside my house and
found it cool and still. By most people's standards my home is
spartan. Although it has three bedrooms and a den, I have only a
small amount of furniture-a scattering of futons, chairs, and
tables. I sleep in the upstairs master bedroom and use the other
upstairs bedroom as an equipment room and staging area for var-
ious adventures and expeditions. In recent years I have spent time
on long backpacking trips, hiked above the Arctic Circle, and fol-
lowed an eclipse in Baja California.
  The absence of furniture is made up for by an abundance of
computers. At anyone time I might have as many as a dozen
machines connected to the Internet at home, depending on what
happens to be plugged in at the moment. Many of the machines
are stacked in one of my closets, and some of them don't even
have monitors, they are simply boxes with processors, memory,
and disks. There are a few PowerLites; a SPARCstation Voyager,
which was Sun's disappointing experiment in entering the
portable computer market; Osiris, a diskless workstation that sits
at the head of my bed that I frequently use as a window into the
Internet; a pair of servers, Rimmon and Astarte, fast Sun com-
puters with big disks that are good for storing data and crunch-
ing numbers; another router; a terminal server; a demo firewall
computer-and the list goes on.
  While most modern offices today connect their computers
with a technology called Ethernet, which was developed at the
Xerox Corporation's legendary Palo Alto Research Center dur-
ing the 1970s, my home computers are linked by fiber optic
cables using a technology called ATM, or Asynchronous
Transfer Mode. An ATM network organizes information differ-
ently than Ethernet. Data are broken down into "cells" rather
than "packets." Cells are generally smaller than packets and
they're all the same size. That means ATM is better designed for
sending video and audio. Moreover, on an ATM network you're
always guaranteed the full speed of the network link; you don't
have to worry about sharing it with your next-door neighbor.
                      DAMAGE CONTROL                          39
A lot of people in the computer and the telecommunications
industries believe that ATM is going to be the wave of the
future. Unlike other network standards it has no single defined
speed, and it's scalable, meaning that it can continue to get
faster as technology improves. My particular implementation is
already fifteen times faster than Ethernet-fast enough to trans-
mit video images that are startlingly clear-much better than
anything you see on today's television sets. Already telephone
and cable companies are laying the groundwork to replace their
existing analog copper wire networks with ATM fiber optic net-
works. Its supporters are confident that by the turn of the cen-
tury ATM data networks will be as easily available in homes as
telephone jacks and cable outlets are today. That's the vision
anyway. I've been quietly experimenting with nitty-gritty engi-
neering details that have to be worked out before any of this
becomes a consumer reality.
   What fascinates me is the power inherent in high-speed com-
puter networks and what you can do with them as opposed to
a single isolated computer. Sun has an advertising slogan, "The
network is the computer." When you get past the hype there is
a powerful kernel of truth there that is implicit in all of the
recent popular interest in the Internet. Single computers are no
longer very interesting, but the distributed computer that is
emerging out of the web is where the future is hidden. As a
result I have orange, white, and beige cables running every-
where. Some of them pass through the walls and some of them
are out in the open. These cables carry computer data as tiny
flashes of light. Imagine switching a beam of light on and off
hundreds of millions of times a second. (As an experiment you
can shine a flashlight in one end of a coil of fiber optic cable.
When you look at the other end you see this razor-sharp star-
like point of light.) One thing is certain-fiber optic cables are
much more resistant to being slammed in doors than standard
copper cables.
   Feeling bone-tired I stood inside my entryway for a moment,
glad to be home but frustrated that I wasn't in the mountains.
Then I turned off the house alarm and padded upstairs to my
bedroom with the intention of taking a nap while I waited for
40                          BREAK-IN
Andrew to arrive. It was a bright sunlit morning and from my
bedroom I could see across the rooftops to Torrey Pines and the
ocean. The room was still. There was no noise from fans or
whirring disk drives. Although there are three computers in the
room I feel strongly that human beings and moving computer
parts don't mix.
   Nothing had changed visibly, but something was odd. Sitting
cross-legged on my bed in front of Osiris's I touched the trackball
and the screensaver that was blanking its monitor gave way to a
field of windows. I noticed immediately that a large rectangle that
sits on the left side of Osiris's screen and that is usually connected
to either the outside world or to Ariel at the Center was com-
pletely blank. It was all white. There was no sign of life-none of
the text that should have been displayed, even if the computer it
was connected to was halted.
   I thought to myself, this is weird, because even though Ariel was
still frozen back at SDSC, Osiris's screen should still register its
presence. I sat on my bed and looked at Osiris again and thought
some more. Nothing really registered. I halted Osiris. I went and
halted its current server computer, Astarte. Then I systematically
froze my other computers as well. My entire computer world was
suspended, as if it had been frozen in midstep.
   I went back downstairs and looked in the refrigerator and real-
ized that there wasn't a lot of food in the house. Not surprising,
because I travel so much of the time. I scrounged around and
found some Power Bars, which would have to do for the
   I returned to my bedroom to try again to analyze the break-in.
My first step would be to obtain some forensic tools to examine
the intruder's tracks. I switched on my new RDI and began
assembling a small toolkit ofprograms that could collect and ana-
lyze data. What I wanted to learn was what files had been read,
modified, or created. It's easy to tell precise times when things
happen on a computer, because the operating system routinely
time stamps any file change. With that information it might be
possible to construct a chronology of the intruder's activities. It is
possible to alter this information systematically, however, and so
I knew it was important not to take any information in the digi-
tal world on faith.
                        DAMAGE CONTROL                             4I
   I now had a handful of frozen computers, in which the
intruder's tracks were hidden as strings of electronic 1's and D's.
My plan was to remove their disks and plug them into a new
computer to do the analysis, for by making the disks "read
only" it would be possible to avoid any danger of accidentally
smudging the data as I explored it. I stared at the portable com-
puter, which was a prototype and might not work. Brand-new
machines tend to have glitches that can be irritating. Maybe I'd
be lucky. If it worked I'd be able to determine what files the
intruder had touched and when. Then possibly I'd also be able
to discover how he had broken into my computers.
   Shortly before noon I called Andrew, who'd arrived in San
Diego several hours after I had and had gone home to drop his
gear. He'd taken an even earlier flight from Tennessee than I had
from San Jose and we were both feeling pretty trashed. We agreed
to meet for dinner that evening to plan a course of action. The
last time I'd spoken with Andrew had been at 2:30 A.M., just
before I fell asleep. He hadn't slept at all the night before, but he
said he was able to doze a little on his flight. Finally, in the early
afternoon I lay down on my bed and fell asleep, only to wake
later in the afternoon still groggy. But I had this sense the next
couple of days were going to be intense and even a short catnap
was an improvement.
   The ghostly image of oki.tar.Z continued to nag at me. What
did it signify? Several years earlier I had helped out Mark Lottor
by reverse engineering the software that is built into Oki cellular
telephones. Normally the programs that control a cellular phone
are hidden in a ROM chip inside the phone. However, most
phones have an undocumented interface to the outside world
that makes it possible to control the phone remotely from a com-
puter. We examined the software carefully and worked backward
from the 1's and D's embedded in the chip to the original com-
mands intended by the software designers. Reverse engineering
software is still controversial, but recent court rulings have gener-
ally held that such work is a legitimate activity. Mark wanted to
be able to control the Oki phone in order to develop a piece of
field diagnostic equipment to sell to cellular telephone companies
and law enforcement agencies.
   Since we had no help from Oki in figuring out how to control
42                         BREAK-IN
their phones, we had to take their software apart to see how it
worked. What we found were lots of undocumented features that
cell phone users have no idea exist. A cellular telephone is really
little more than a radio with a tiny personal computer, so when
we looked closely at the Oki's software it came as no surprise that
it had obviously been written by truly clever hackers.
   With commands that can be punched into the keypad of an
Oki phone, it is possible to obtain all kinds of diagnostic data on
how the phone is behaving, such as its signal strength, that are
quite useful for phone technicians. Many brands of cell phones
also happen to function just as well as the Oki as cellular tele-
phone scanners. Few people realize that if they know the right
buttons to push on their cell phone keypads, they can easily lis-
ten to all the phone conversations that are going on in their
neighborhood-a trick which is, of course, a violation of the
Electronic Communications Privacy Act. But since there is no
privacy at all in today's cellular telephone system, illegal eaves-
dropping on cell phone calls has become a popular pastime.
    In 1992 I testified before a Congressional hearing held by
Representative Edward Markey on the existence of this undocu-
mented cell phone capability. After the committee chairman
granted me special immunity, I took a brand-new AT&T cellular
phone-actually the same Oki phone but relabeled and sold by
AT&T, still in its shrinkwrap packaging-assembled it and
pushed a series of buttons. Soon the committee was listening to
cellular phone conversations from all over Capitol Hill.
   Afterwards a bulky, middle-aged FBI agent came up to me and
said, "You have congressional privilege now, but don't ever let me
catch you doing that outside this room." His comment con-
firmed one thing I've noticed in working with the FBI: these guys
have no sense of humor.
    Oki.tar.Z not only suggested a motive for our break-in, but it
also hinted at who the perpetrator might be. A few months earli-
er in October and November, someone had repeatedly broken
into Mark Lotter's computers in an effort to steal the same Oki
cellular software that had been pilfered from Ariel.
    Mark was in the process of setting up a new cottage business.
The Internet was booming, and he had found there was a ready
                       DAMAGE CONTROL                           43
market for publishing pages on the Internet's rapidly expanding
World Wide Web. Network Wizards was therefore creating
"catalog.corn," an inexpensive web site allowing people to dis-
play catalog information or whatever else they wanted to com-
municate. The Web, originally developed as a scientist's research
tool by a computer programmer at CERN, the physics research
center in Geneva, Switzerland, had emerged almost overnight as
a convenient vehicle for permitting electronic commerce on the
   In addition to his Network Wizards fileserver, Mark's home
Ethernet network also supported two other computers. Lile,
Mark's friend, had created Art on the Net, a virtual artists'
gallery housed on a donated Sun workstation that enabled a
new generation of digital artists to exhibit their works. Another
Sun on his network had been donated as a web site for the
League for Programming Freedom, a organization of hackers
dedicated to Richard Stallman's crusade to create a world of
free, shared software.
   For several weeks in early October, Lile had been experiencing
strange behavior with her electronic mail account at the com-
mercial Internet service provider Netcom. She would attempt to
have her mail forwarded to her art.net Sun from Netcom only to
find a short time later the forwarding file had disappeared. She
complained to Netcom, but the telephone support person told
her it couldn't possibly be a security problem, explaining, "We
haven't had a break-in for three weeks."
   Then, one Saturday morning in mid-October, Mark woke up
and came downstairs to make himself an espresso. He went to his
computer to read his mail and was sitting near his fileserver when
for no apparent reason it began to make a prolonged grrrrrrr
   That's odd, he thought to himself. The computer, which was
linked by a high-speed T-l connection to the Internet, was sup-
posed to be idle. When he connected to the machine he saw that
a long listing of all of his files was in the process of being dis-
played. He looked further and saw that someone was running a
root on his computer.
   His first thought was maybe this was the result of some
44                          BREAK-IN
unusual program he wasn't familiar with. Unix computers have
lots of small programs called daemons that constantly run in
the background, performing housekeeping tasks. Then he ran a
program called netstat, which gives detailed information about
what's happening on a local computer's network connection.
He could see that someone was connected to his machine from
Lile's art.net Sun.
   But Lile was sitting across the room from him at her own com-
   ''Are you telneting to my computer?" Mark asked, referring to
a utility used for connecting to a remote computer across the
Net. She wasn't.
   His alarm grew until he began to panic when he saw the per-
son logged into his machine begin to tar a group of his files
together. Seconds later the cracker began to use ftp-file transfer
protocol-a common Internet file transfer utility, to move the
tarred file to Netcom.
   Mark was horrified. "What am I going to do?" he said to Lile,
who had joined him in watching in disbelief as the huge file was
copied out of his computer. He looked around his apartment and
realized his quickest defense was to take himself off the Net. He ran
and pulled his T-1 data cable out of the wall.
   Mark and I chatted by phone later that day. He had examined
the tarred file after he had pulled the plug from the network and
discovered that someone was definitely trying to get the Oki tele-
phone software that we had modified for his cellular telephone
diagnostic system. He was able to determine they had been
unsuccessful in getting anything of real value, only a small chunk
of the file. He warned me to be on guard, and a short time later
Andrew and I had seen some probes against our computers which
he easily fended off.
   The next day Lile and Mark headed down Highway 17 to
Santa Cruz to visit a hot tub place near the campus of the
University of California at Santa Cruz. While they were driving
 through the Santa Cruz mountains Mark's cell phone rang.
   He answered, and a voice said, "Hi." Mark recognized the
 caller immediately as someone he knew only slightly but who had
links to the computer underground.
                             DAMAGE CONTROL                             45
I       "I haven't forwarded my phone and I don't give out my cell
:\   phone number," Mark said. "How did you get my number?"
        "Let's just say I came across it somehow," the caller answered.
     "I just wanted to tell you I know who broke into your computer
     yesterday. "It was Mitnick and his friends, and they were really
     pissed they didn't get what they wanted."
        The name Kevin Mitnick was a familiar one to Mark, as it
     was to anyone who followed the history of the shadowy world
     of the computer underground. Mitnick had grown up in the
     San Fernando Valley in Southern California during the 1970s
     and had made the transition from the hobbyist world of phone
     phreaks who tampered with the telephone system, to computer
     crackers who used networks to break into computers. There
     seemed to be one significant difference between Kevin Mitnick
     and the thousands of teenagers who seemed to be imitating
     Matthew Broderick in the movie %r Games. Mitnick seemed
     remarkably incorrigible. Barely thirty-one years old, he had
     been arrested five times already, going back as far as 1980 when
     he was only seventeen. Mitnick was currently on the run from
     a variety of law enforcement agencies including the FBI.
        John Markoff, a New York Timestechnology reporter both Mark
     and I knew well, had co-authored a book called Cyberpunk, about
     computer crime including Kevin Mitnick. He had also written an
     article about Mitnick in July 1994, saying that he had outfoxed
     federal and state officials for more than a year. The article reported
     that Mitnick was a suspect in the computer network theft of soft-
     ware from as many as a half-dozen cellular telephone companies.
        Mark remembered that several weeks before his Saturday morn-
     ing attack, someone he knew as an old friend of Mitnick had called
     and said he wanted to buy Network Wizards' cell phone software,
     but he also wanted source code, the original programmer's instruc-
     tions that would make it possible to further modify the phone's
     functions. Although Mark declined to sell the source, Mitnick's
     friend had stayed on the phone for more than an hour wheedling.
        There were no more attacks over the weekend, but during the
     next two weeks, the intruder continued to be a constant pest,
     breaking into Lile's art.net machine and the League for
     Programming Freedom computer repeatedly, leaving Trojan
46                           BREAK-IN
Horses and back doors.
   From time to time he would engage her in chat sessions, using
a command called talk that permits two users of a Unix system to
type back and forth to each other in real time at their keyboards
over the Internet.
   "Why don't you just give me the software?" her screen read one
day. "I'm going to get it anyway."
   He also asked for an account on her system, again telling her
he would get one anyway. Lile offered him one of her digital
artist's virtual studios, but he wasn't interested. He said if she gave
him an account he would reveal who he really was.
   "I hope you aren't upset with me," he typed
   Mark was sitting in the room at the time and he coached her
on how to respond. They tried to get him to give away little bits
of information about himself, but with little success.
   Finally in early December, the invader called Mark on the tele-
phone directly to try to persuade him to give him the software.
   "Do you know who I am?" he said. "I want your code," and
proceeded to ask Mark if he was upset about the break-ins to his
   Mark responded that he wasn't, and explained that he had a
different philosophy of computer security: if anyone managed to
get inside his system, it would just alert him he needed to
strengthen his defenses.
   "Then I'll just continue trying," the caller said. Mark asked him
about why he was so intent on getting the source code to the Oki
cellular phone. The anonymous voice responded he wanted to be
invisible in the cellular telephone network and he believed that if he
was able to alter the behavior of his phone it would put him beyond
the reach of cellular telephone tracking and surveillance gear.
   They talked three times. The first two phones calls were short,
but the third went on more than forty-five minutes during which
the caller asked, "You're not taping this, are you?"
   Mark said no, but then decided it would be a good idea. He
quietly moved across the room and hit the "record" button on his
answering machine.
   The caller knew who I was and knew that I had helped Mark
with the cellular telephone project. He seemed to be probing for
 more information about me:
                       DAMAGE CONTROL                         47
Caller:   God, so you actually, uh, so he wrote that dis-
          assembler, I see ...

Mark:     Yeah.

Caller:   Why did he write it? For you, or just he hap-
          pened to just write an 8051 disassembler?

Mark:     Um ... urn, I don't r-emernber just why he wrote,
          it. He just hacked it up one evening.

Caller:   Shit. just one evening? That's imp ...

Mark:     Uh , I think it only took him an hour or two,

Caller:   No way!

Mark:     (sigh)    (laughs)

Caller:   Are you serious?

Mark:     Uh huh.

Caller:   That guy's a wizard. He should work for your
          company, Network Wizards.

Mark:     Uh        he has better things to do.

Caller:   In San Diego still I assume?

Mark:     Um, sometimes.

Caller:   And Los Alamos?

Mark:     Sometimes.

  After Mark hung up he called Markoff and played him the
tape to see if he recognized the voice. The reporter had never
formally met Mitnick but had heard his voice on the telephone
and on tapes several times. He said it sounded like him but he!
wasn't positive.
  Mark next called Jonathan Littman, a Marin County freelance
writer who was writing a book about the computer underground
48                          BREAK-IN
and who was rumored to have a secret channel to Mitnick. Mark
played him the tape and said, "Do you recognize this voice?"
  Littman started to laugh. "Of course I do. It's Mitnick."

The possibility that Mitnick was the culprit in my break-in as
well as Mark's was an intriguing thought, but one that really did-
n't lead anywhere useful now, so I put it aside. Lots of people in
the computer underground and elsewhere knew that I had
worked with Mark on cellular telephone software. At this point
the first thing I wanted to do was collect data and find a way to
secure our computers as quickly as possible. I began scanning the
first of the disk drives that I'd extracted from the computers that
had been frozen, using the software tools that I'd gathered. I
wanted to find all files that had been read, written to, had their
dates changed, or new files that had been created going back to
December 21 when I had left San Diego. I sat for a long time in
front of the PowerLite. Somewhere in the morass of data I was
almost certain to find a clue or a set of clues. No one can be per-
fect at hiding their presence.
   I was also searching for Trojan Horse programs. These are pro-
grams often left behind by electronic intruders. They can wake
and silently do any number of secret or destructive things. They
masquerade as familiar software but could be written to eaves-
drop on you, destroy data, or provide a convenient back door
security loophole. One way to protect your computers against
this kind of tampering is to take a digital snapshot of all the pro-
grams that are on your computer-operating system programs,
utilities, communications tools, everything. Later you can tell if
any files are altered by comparing mathematically generated sig-
natures of the files on the suspect disk against those of your orig-
inal safely stored copy.
   About nine that night Andrew and I met for dinner at a place
near campus called Pizza Nova. Andrew is an example of an
Easterner who has adapted remarkably well to the California
beach scene. With a head of shoulder-length blond hair, a promi-
nent nose, and intense blue eyes, his standard outfit is shorts,
T-shirt, and sandals. Andrew is also well known for the fact that he
                             DAMAGE CONTROL                             49
     really doesn't like to wear shoes at all, a trait that can sometimes
I    cause trouble when we try to eat in restaurants. He has the hacker-
I'   like quality of being able to focus on complex problems for extend-
~    ed periods of time, sometimes aided by several liters of Mountain
     Dew. I get frustrated with him occasionally because he jumps to
     conclusions and acts too quickly rather than thinking through the
     consequences of particular action. But he has a good intuitive grasp
     of the inner structure of the Internet and we work well together as
     a team-and I find him a pleasure to work with.
        Over dinner we talked about things that needed to be
     explored. We agreed that for now, it was necessary to concen-
     trate on actually addressing all the possible vulnerabilities in my
     network. One thing that puzzled both of us was that the intrud-
     er had played around with XNeWS, a PostScript-based compo-
     nent of the operating system that permits you to draw images
     on your workstation or on a remote computer. PostScript is
     more widely used as a printer language, giving programmers a
     set of commands to tell printers where to draw lines, place
     printed characters, and shade areas. Could that have been a vul-
     nerability? Maybe the intruders had found a design error in
     PostScript that would allow them to use it to remotely take con-
     trol of a computer. I gave Andrew a set of tasks and took a set
     for myself. We parted with the agreement that we would recon-
     vene at the Supercom-puter Center the next day.
        I drove home feeling emotionally drained and even more
     exhausted; in fact, things looked bleak. We had a tantalizing set
     of hints about how my systems had been attacked. But there was
     no certainty we would be able to make any sense of them, and
     even if we were able to reconstruct the crime there was no likeli-
     hood that we would be able to follow a trail back through the
     Internet if our attackers had really covered their tracks well. It was
     eating at me, and forcing me to think about things that had been
     bothering me that went far beyond this particular break-in.
        When I got home, Julia called. She was at Toad Hall, and we
     both had been having a difficult time.
        ''I'm here and I feel like I'm stuck here," I told her. We talked
     about the break-in for a while, and then we talked about what she
     had been doing in San Francisco.
50                          BREAK-IN
   ''Mter John came back, things went well at first," she said, "But
things are much more tense now."
   It was clear their relationship hadn't been working since I'd met
Julia, and it didn't look like anything was changing. Onour first
wilderness trip together three years ago we'd gone snow camping
in the Desolation Wilderness near Lake Tahoe, and she'd told me
of her misgivings about her relationship with John. She wasn't
happy and wondered aloud whether or not she should stay with
it and try to make it work. We'd talked late into the night and
she'd told me she thought she should keep trying out of a sense
of loyalty to her partner.
   Now three years later I could tell she knew the relationship
was bad for her, but she didn't seem to be able to get out of it.
I knew that she had been trying to end it for some time, but she
found familiar things comforting and difficult to part with. It
was all making her unhappy and depressed. It wasn't the first
time I'd seen her feel this way-a year ago she'd gone to Nepal
with John-but he left and she became involved with another
man, an American she met while traveling. Their relationship
went on for six months, but it ended over her unwillingness to
leave John.
   It was as if there were two Julias. One was a strong, indepen-
dent, and adventurous woman who was trying to find what
would make her happy and content. But there was another per-
son as well, who was hamstrung by fear and feelings of inade-
quacyand insecurity. I'd seen her grow stronger and more inde-
pendent since we'd met, and she was better able to see what was
damaging to her, but she hadn't been able to bring herself to
make a final break with John.
   We talked more about my problems in San Diego. I was
unhappy to be here at all, given my skiing plans, and the more I
looked at the problem the more apparent it became that it wasn't
going to be simple and that I might be wasting my time in a futile
search to find out how the break-in had taken place.
   Over the past few months I'd already been feeling burned-out
by my dealings with NSA. I'd had enough of dealing with com-
puter security for a while, and I'd been looking forward to skiing
and working on other problems. But now, as I told Julia, I felt
                       DAMAGE CONTROL                           5I
stuck. I was being pulled back to deal with computer security, but
without the resources I needed.
   "This is the last thing I want to do right now, but I can't walk
away from it," I said.
   "You sound awful, Tsutornu. I'm worried about you," she
replied. She offered to come down and spend time with me.
   But I said no, she had enough to deal with without having to
come down and comfort me. I was in a foul mood and wanted to
focus without interruption, in order to be done with the threat as
soon as possible.
   "Get some sleep tonight," she finally said. "Between you and
Andrew, the two of you will be able to make some progress
tomorrow. "
   It seemed like there wasn't any other option. We said good
night, agreeing that we would talk again soon.
   Just before going to sleep I remembered that I had forgotten to
listen to my voice mail since returning to San Diego. I played
back a string of messages that had been left for me at my SDSC
office. I listened to four or five routine messages, continually
pressing the phone's keypad to zap them out of the system's mem-
   Then I heard something that made me sit up in my bed.
   "Sent December 27 at 4:33 P.M.," said the prim, electronic
female voice.
   Immediately came another voice. It sounded like someone who
was faking a passable Australia~ cockney accent. There was no
mistaking the message.
   "Damn you," said my caller. "My technique is the best. My
boss is the best, damn you. I know rdist technique, I know send-
mail technique, and my style is much better."
   Rdist and sendmail were two garden-variety computer network
attacks involving long-known computer system vulnerabilities.
This could only be my intruder, calling to taunt me.
   "Damn you, don't you know who I am?" he continued. "Me
and my friends we'll kill you."
   Then it sounded like he turned his head away from the phone
in order to sound like another person's voice: "Hey boss, your
Kung Fu's really good."
52                         BREAK-IN
   "That's right," concluded my caller III the same Australian
accent. "My style is the best."
   This time I saved the message. Then I lay back on my bed and
stared at the ceiling. This was getting personal, and whoever this
was had obviously become pretty cocky. I don't need this, I
thought to myself. If it hadn't been certain before, now it was.
Someone was clearly challenging me.
i                         CHAPTER
I                                             4
    We frequently call things that we don't understand in the world
    complex, but that often only means that we haven't yet found a
    good way of thinking about them.
      Throughout my life as a scientist I've been absorbed with
    exploring and understanding complexity, and I've found that
    although it might seem that nature could have come up with a
    complicated way of making things work, there is almost always
    a very elegant and simple explanation underlying any phe-
      That basic outlook has underscored much of my work in fields
    as diverse as biology and physics, and problems involving com-
    putation, which I have focused on for more than a decade and
    since 1989 as a senior fellow at the San Diego Supercomputer
    Center. How does the physical world compute its answers?That
    may seem an impossibly vague question, but it lies at the heart
    of a radical approach to much of modern science. For example,
    what ways are there to find a leak in a bucket? On the face of it,
    a computer would have no simple method of finding the point
    of the leak. It could iterate over the entire surface of the bucket,
    point by point, until it found the hole. But there is a better, sim-
    pler solution: you fill the bucket and let the water do the work
    of finding the leak.
       It was thinking about such questions regarding the nature of
    computation that had consumed the legendary physicist Richard
54                         BREAK-IN
Feynman toward the end of his life. I began taking Feynman's class
on the physics of computation when I first entered Caltech as a
freshman in 1982, and spent more time with him during the sum-
mer of 1984 at Thinking Machines, a start-up supercomputer
company in Cambridge, Massachusetts. His outlook influenced
my own thinking tremendously. Feynman had a remarkable abili-
ty to see the world clearly, and not to be misled by commonly held
preconceptions. Throughout my career I have sought to emulate
Feynman's approach to science, and I think I have also been helped
in gaining this independent perspective through being a scientist
and by having grown up between two cultures.
   I was born on October 23, 1964, in Nagoya, Japan. My parents
grew up in Japan, where they lived through the war. My father,
Osamu, was trained as a biochemist, and my mother, Akemi,
began work as a pharmacologist. Together they embarked on their
careers as research partners, specializing in the study of biolumi-
nescence. In the 1960s the leading institution for bioluminescence
research was Princeton University. This period was a wonderful
time to do basic research in the United States, and my parents came
here when my father accepted a position as a Princeton research
faculty member. My mother took time off from her career to raise
me and my younger sister, Sachi.
   Although I have early memories of traveling back and forth
between the United States and Japan, the first times I can remem-
ber clearly are the years growing up in Princeton. I particularly
recall learning English as a second language in kindergarten and
first grade.
   Being raised by two scientists permanently. shaped my
approach to the world. My childhood was spent between my
mother's kitchen and my father's lab. From my first steps my fam-
ily encouraged me to be curious. I was provoked to ask questions,
to which I never received "because" replies. My parents' response
was often to suggest an experiment through which I could deter-
mine for myself the answer.
   The value of experimentation was taught to me even in the
most commonplace occurrences. Once at dinner I dropped a
mushroom on the floor, and when I went to pick it up and eat it
my father said, "It's dirty."
   "I can't see any dirt here, " I retorted.
                        THE REAL WORLD                            55
  The result of this discussion was that my father drove me in to
his lab, so that we could examine the dirty mushroom more
closely under a microscope.
  I was, in general, an argumentative child, though my parents
tolerated a remarkable amount from me. I was so quick with
rebuttals, even at elementary school age, that my mother threw
up her hands one day in exasperation and said, ''Are you going to
be a scientist-or a lawyer-when you grow up?"
   In the sixties and seventies Princeton was the quintessential lib-
eral academic community, but I was still an outsider, even though
the university had a large Asian population. During this period I
returned to Japan frequently, and even lived for the better part of
two years there-just enough to ensure that I retained a feeling
of being slightly apart from both cultures. I spent my entire fifth
grade in school in Nagasaki. In Japan both Japanese and English
are taught, and it was interesting to come to know the Japanese
view of America after I had experienced it personally.
   Because my father's research involved studying bioluminescent
jellyfish, I passed many of my summers in Friday Harbor in the
San Juan Islands in Washington, where he conducted field
research at the University of Washington's marine biological lab-
oratory. I was in my element, let loose with lots of other bored
children of academic families. These summers gave me free time
to help out in my father's laboratory trying to find something
useful to do when I wasn't getting into trouble and roaming the
island wilderness. The cool climate, carpets of douglas firs, and
crystal-clear tidepools were a wonderful counterpoint to the more
civilized and hot and muggy suburban Princeton summers.

When I was twelve years old-because I had skipped several
grades I was already in my first year of high school-I began
spending less and less time at home. I wasn't getting along par-
ticularly well with my parents by then, and I ended up being at
the University more and more of the time.
   At the time a friend of mine had a job with a psychology pro-
fessor in a lab that was doing neuropsychological research, and
he helped me get work there as well. He was working on trying
to get a data acquisition system to function. It was basically a
56                          BREAK-IN
programming job involving a DEC PDP-11134 computer, which
at the time was the standard piece of laboratory computing
   The first time I'd seen a computer was in kindergarten. One of
my classmate's fathers, who worked for RCA's Sarnoff Laboratories,
brought a computer to school for some kind of a show-and-tell ses-
sion, and although I didn't get to play with it, it was definitely
something I remembered and was intrigued by. I clearlyrecall that,
even at this first encounter, I considered these machines as tools
that would help me solve problems.
   My real immersion with computing, however, didn't begin
until I was ten when, through friends, I stumbled across a quirky
and informal Princeton computer club known as the Resistors.
Resistors was an acronym for "Radically Emphatic Students
Interested in Science Technology and Other Research Studies,"
and consisted of an anarchic group of teenagers (the average age
was probably 15) that met in the E-Quad, Princeton's four-story
school of engineering. The first generation of Resistors were
influenced by people like Ted Nelson, the visionary social scien-
tist who wrote the book Computer Lib/DreamMachines and who
was to become the pied piper of hypertext. I was actually a mem-
ber of the second generation of the Resistors, and what the group
gave me was an easy way into the world of computing-in effect,
my own personal computer in the era of minicomputers and
mainframes. But I was also a little bit of a loner, and I never had
any really close relationships with the other Resistors.
   The group was actually a result of the same urge that would
drive a similar hobbyist club known as the Homebrew Computer
Club, which emerged several years later in Silicon Valley. While
its members were older, in their early twenties, for the most part,
they also were a product of the availability of the first inexpensive
microprocessor chips, and their passion to have their own com-
puters led directly to the explosion of the personal computer era.
People like Steve Wozniak and Steve Jobs created Apple
Computer in the hothouse technological culture that emerged
around the Stanford campus in the mid- to late seventies. Other
Homebrew members like Lee Felsenstein went on to design both
the Sol and Osborne 1 machines.
   The Resistors grew up in the thrall of an earlier computer age,
                        THE REAL WORLD                            57
with a distinctive East Coast flavor. The mainframe world had
emerged during the fifties and sixties at IBM, followed by the
minicomputer era of the seventies created by companies like
DEC, Data General, and Prime. The minicomputer was the
underpinning of the timesharing epoch of computing. A product
of MIT hacker culture, timesharing operating systems permitted
more than one person to use a computer simultaneously. The
trick was to slice computing tasks up into tiny pieces and then
have the central processing unit of the computer jump quickly
from one task to another in a round-robin fashion. This scheme
made computers tremendously more productive and also
expanded the power of computing to a vastly broader audience.
It was timesharing that also enabled young hackers like me to get
access to powerful computers.
   AT&T's particular contribution to the computing revolution
had been the Unix operating system, a program that was developed
by two Bell Laboratories computer scientists, Dennis Ritchie and
Ken Thompson, in the late sixties. Operating systems are a combi-
nation traffic cop, secretary, and waiter inside a computer. They are
the programs that perform all of the basic housekeeping operations
and respond to user requests and commands in addition to orches-
trating the delicate ballet that goes on between all the different
components of a computer system. Operating systems also provide
a computer with a distinct personality. Think of them as a language
for speaking directly to the computer hardware.
   Ritchie and Thompson, who had been aghast at a cumbersome
Pentagon-funded operating system development project called
Multics, created Unix as an alternative, and it quickly became the
programmers' tool of choice for the ragtag army of hackers who
were forming a computer culture at various universities and com-
panies around the country. Like thousands of college students at
the time, I grew up a child of the Unix revolution.
   Unix was a system-created by stripping down many of the
features of the big mainframe world and adapted to the capabili-
ties of minicomputers and workstations-unlike the personal
computer world, in which operating systems like CP 1M and MS-
DOS and Apple DOS were built from the ground up. As a result,
my generation of hackers expected computers to have certain fea-
tures that personal computers didn't have, and in some cases still
58                         BREAK-IN
don't have, until more than a decade later. Unix computing con-
cepts like multitasking, hardware memory management, and
portability were all part of a gospel that I learned while I taught
myself Unix programming between the ages of ten and fifteen. It
only made sense that computers should be able to do multiple
things at the same time, even for a single user, and utilize hard-
ware memory management, which ensures that a poorly designed
program doesn't leave the memory space that has been reserved
for it and trample other programs.
   I also grew up having no idea that there was any way to use
computers other than connected to one another in networks. I
first came in contact with the ARPAnet, the Pentagon-funded
predecessor to the Internet, in 1976. It was a wide-open, if tiny
community-there must have been a grand total of one hundred
computers on the entire network-and I loved to explore it.
   What I didn't do was spend my afternoons and evenings break-
ing into computers, a fad that teenagers took up almost a decade
later. When I was traveling around the Net in the mid-seventies
there were almost no locks, and everything was shared. At various
places around the Princeton campus were public dial-up termi-
nals that would allow you to sit down and access the entire Net.
I would play games, chat with people, and hop around to places
like MIT, Carnegie Mellon, and Stanford.
   Although my first programming language was BASIC, a com-
puter language created for education in 1969 at Dartmouth, in
the course of playing with it I soon realized I could escape from
the narrow confines of a limited programming language to the
Unix command shell. The shell-basically the computer's soft-
ware control panel-broadened my horizon and gave me access
to all of the computer's resources, as well as access to the network
world beyond. After being confined to the tightly controlled
world of Basic, the Unix shell was a little like being on the cap-
tain's deck of the Starship Enterprise. Reveling in my new freedom
 I learned to program in C, a language that had been developed
 by the same crew of Bell Labs hackers who had invented Unix. C
was liberating for me. As a language it is obtuse, but once it is
 mastered, it is remarkably powerful and flexible. C also gave me
 a skill that put me in great demand, even as a teenager.
    A computer science graduate student named Peter Honeyman
                        THE REAL WORLD                           59
offered me my first Unix account at Princeton on the electrical
engineering and computer science department's DEC computer.
Although it was later taken away when the computer became
overloaded with legitimate students, it was my original entry
point into a world that was soon to become my passion. The
PDP-ll/45 was a flaky machine, and Honeyman was obviously
smart enough to see that I and another high school student, Paul
Rubin, were an untapped source of cheap labor. We soon became
the computer's administrators and caretakers.
   Occasionally I got in trouble for my computer use. I remem-
ber discovering that I could issue commands that would move
the head-arm of the magnetic disk drive on our high school mini-
computer back and forth in a rapid fashion. The drive was a stan-
dard 14-inch IBM frisbee-style monster, whose head for reading
and writing magnetic data was controlled by a stepper motor that
could be commanded to move in and out and read and write on
any of the drive's 203 cylinders. After I successfully moved it to
cylinder 0, and then to cylinder 100, I began to wonder, "Gee,
what happens when you try to position it to cylinder HEX FFF?"
That would be the equivalent of cylinder 4095-unfortunately
one that didn't exist.
   I issued the command and heard, Rrrrrr. Crunch. That was the
end of the disk drive. It was a useful lesson, and for me it effec-
tively put to rest one of the canons of computing that are always
taught: "Don't worry, you can't hurt the hardware."
   High school increasingly became a sideshow in my life, as I
spent more and more of my days around the university. At one
end of the spectrum I was fascinated with physics, because it
explored the fundamental principles underlying everything in the
universe, and at the other end, biology, where very simple princi-
ples create very complicated systems. In one case you can try to
simplify, while in the other, there's no chance of simplifying. I did
read in other fields like psychology and geology, but I wasn't
intrigued by them, for they were more like botany-in my mind,
the art of categorization involving little analysis or understand-
   While I benefited from this sampling of the academic disci-
 plines, it was my computing skills that made me an integral part
 of the university scene. In 1978 Princeton was facing a growing
60                          BREAK-IN
problem in the proliferation of minicomputers. The school's com-
puting center had earlier tried to maintain a monopoly over all aca-
demic computing. It had, in effect, told the various departments,
"Since we provide this service, we want you to use our computers."
And like any monopolist, they charged outrageous rates. However,
Princeton's other departments discovered a loophole that would
free them from having to deal with the computing center-special-
purpose applications. The various departments all managed to jus-
tifY buying their own individual minicomputers to carry out odd-
ball special-purpose projects that they dreamed up.
   The astronomy department had been able to wangle its own
DEC PDP-11160 and wanted to run Unix, but there was no one
on staff who knew anything about it. Since I was already hang-
ing around much of the time begging, borrowing, or stealing free
computer time, they asked me to come and install some special-
ized hardware to help them read a particular magnetic data tape.
   Through that project, I became the department computing
wizard at the age of fourteen. After I wrote a special device driver
for them, enabling the computer to reliably talk to some esoteric
disk drives, I was invited to work part-time by a young assistant
professor, Ed Turner. The tradition in the department then was
that computing was the responsibility of the newest faculty mem-
ber. Ed not only hired me, but introduced me to a remarkable
world outside of my claustrophobic high school existence.
   Looking back on it, taking that job was one of those watershed
events that helped me define who I was at a very young age. It
was also an incredible amount of fun, and it gave me access to
some of the world's best toys.
   I was around the astronomy department constantly throughout
the rest of my time in high school, mostly helping with comput-
erized image-processing work. While I learned something about
pattern matching and a great deal about systems programming,
the job also permanently shaped my attitude toward computing.
Trying to solve problems in astrophysics and astronomy con-
vinced me that, if a machine doesn't do what you want it to do,
then reprogram it to do the right thing. Early on I learned that a
machine does just what you tell it to, and nothing more.
   My responsibilities evolved to include designing special hard-
ware for the astronomy department, and in my senior year I was
                        THE REAL WORLD                           6I
able to help develop a data-storage system for capturing experi-
mental data from a NASA-funded missile launch in which the
department was participating. My task was to help design the hard-
ware that would actually collect the bits of data from the flight at
the White Sands Missile Range in New Mexico. The problem
involved downlinking images from cameras aboard the missile,
storing them on videotape recorders, and then trying to convert the
data into digital form for computer storage and analysis. It proved
to be a successfullaunch, which brought back data in the ultra-
violet range of the spectrum from the fringes of space.

Being around an older community of students and faculty at such
a young age, and being essentially precocious, also ultimately got
me in trouble. Because I had skipped ahead two grades, I never
graduated from junior high, and in the fall of 1980, I was a fif-
teen-year-old high school senior.
   Despite uninspiring grades, I received stellar college-placement
test scores, and I was accepted at Carnegie Mellon early in my
senior year. I was disappointed, however, that my other choice,
MIT, rejected me, even though I had recommendations from a
number of faculty at both Princeton and MIT. Throughout my
high school years, my attitude had been that as long as I was
doing something sufficiently "intellectually" interesting, grades
wouldn't matter.
   Bored and increasingly resentful of what seemed like silly aca-
demic rituals, I would occasionally provoke my teachers. In one
English class we were given a list of words that we were required
to use in essays to be written over the course of the semester. We
were supposed to use the words properly in context, and for each
correct use we would be awarded a certain number of points
counting toward our final grade.
   However, it occurred to me that there was a simpler solution,
which I dispatched in a single ten-minute assignment. I wrote a
story about a silly English class in which the teacher assigned a
list of words, which I included verbatim, as a list. I demanded
credit for the entire list, and decided that since I now had enough
points to pass I wouldn't show up for the rest of the quarter.
   Such stunts didn't endear me to the faculty.
62                          BREAK-IN
   I also didn't fit the stereotype of a computer or science nerd in
high school. Although I didn't go out for high school teams, I did
love doing all kinds of athletic things on my own. I became an
avid cyclist, racing bicycles with a local racing club called the
Century Road Club ofAmerica. I subsidized my relatively expen-
sive cycling gear by occasionally building bicycle wheels for a
local bike shop, Kopp's Cycles. During the winters I began to
develop a passion for cross country skiing in the rolling New
Jersey countryside.
   During the little amount of time I did spend at school, I hung
out with a small group of friends, one of whom we referred to as
"the terrorist." He was actually a gifted classical pianist, and
although the administration would occasionally suspend him,
they couldn't really take any more serious steps because they
needed him to play at school functions.
   He was quite notorious, and his pranks were always dramatic:
the stairwell was thermited. The toilets appeared in the middle of
the football field. The public address system was once complete-
ly fried because he considered it to be a tool for propaganda. It
literally smoked.
   At the time of the latter prank, I was five-hundred kilometers
away, visiting Carnegie Mellon for a college interview, and so had
what I knew was the perfect alibi. But I was a known trouble-
maker, and when I walked back into the school the principal
came up to me and said, "You! You did it!"
   Later the administration figured out that the PA system had
been destroyed when 120-volt alternating current was fed into
one of its call buttons with a timer. They discovered this only
after they had swapped out the boards and the whole system blew
up again, because the timer was on a 24-hour cycle.
   The school paper was planning to run an article on this inci-
dent, which we knew was full of technical errors. Late one after-
noon we went to the newspaper office, took the article from the
editor's box, revised it, and then put it back. The finished piece
turned out to be quite technically accurate, although no one in
the school appreciated how it ended up that way.
   "The terrorist" eventually went to Yale, but during high school
he never managed to be regarded as anything but a vandal. For his
 own part, he thought of his actions as political crimes. He was
                        THE REAL WORLD                            63
quite well-developed ideologicallyas a high school student. I'm still
not sure where any of my group would have fallen on the political
spectrum. Probably antiestablishment-any establishment.
   In my own case, I don't know what the final straw was: it might
have been grades or it might have been insubordination. But one
night during my senior year, I came home for dinner to find that
my parents had received three letters from the school. Two of
them were commendations for my having won a local math and
physics contest, the other was a notice of my expulsion from
Princeton High School. Giving up on me as a lost cause for his
educational system, the principal had told me earlier that week,
"Don't come back, you're persona non grata. If we find you here
we'll have you arrested."
   My response was, "Okay, fine! I didn't come here when I was a
student. What makes you think that I might want to come back?"
   When I was kicked out of school, Carnegie Mellon rescinded
its offer. They said they would hold my place for the next acade-
mic year, giving me another chance to graduate. I ended up in a
shouting match with their recruiting officer and told him, "Don't
   Shortly afterward, my parents joined the marine biological lab-
oratory in Woods Hole, where my father had accepted a research
position. I still had my job in the astronomy department, and my
friends were still in Princeton, so I found myself shuttling back
and forth between New Jersey and Massachusetts.
   Despite my experience in high school, I had always had aca-
demic aspirations that focused increasingly on physics, and so I
applied to the University of Chicago, Johns Hopkins, and
Caltech. When I had first begun the college application process,
I had been under the impression that Caltech was too intense,
a little like drinking from a fire hose, and so I hadn't applied
there. But I had worked for Jim Gunn, a bright young
Princeton astronomer who had begun as a astrophysics profes-
sor at Caltech, and after my rejection by Carnegie Mellon, he
and a couple of others in astrophysics were intent on seeing that
I still went to college, and thought that Caltech might actually
suit me very well.
   With my test scores and recommendations from people like
Gunn I was admitted to Caltech for the fall of 1982.
64                         BREAK-IN

In the summer of 1982, at age seventeen, I traveled to Southern
California, planning to study physics and biology when college
began a month later. I'd been to Caltech a number of times
before, through my work in the Princeton astronomy depart-
ment, and I took a room in a house just off campus. Compared
to Princeton, it has always struck me as a tiny campus. Located
in Pasadena, tucked up against the San Gabriel Mountains, it felt
claustrophobic; it was too smoggy to bicycle in Los Angeles, and
as I didn't have a car, I had no means of escape. It didn't escape
my notice that the school motto ''And ye shall know the truth,
and the truth shall make you free," from John 8:32 in the Bible,
was the same motto as the one employed by Central Intelligence
   For my application interview the previous spring, a Caltech
faculty member named Jerry Pine had visited me in my office at
Princeton. Pine, a high-energy physicist who had made the tran-
sition to biology, had suggested I come to California before
school started to work during the summer on a project being led
by another Caltech faculty member, Geoffrey Fox, a physicist,
who had heard from Gunn and others about my reputation as a
hacker. Fox was in the early stages of designing a new kind of
massively parallel computer known as a hypercube. It was a pow-
erful, novel computer architecture that was in keeping with the
trend toward breaking down complex problems into small com-
ponents and computing them simultaneously. Parallel comput-
ing, then being pursued by researchers and companies around the
country, would later lead to dramatic increases in the speed of
computing, and transform the supercomputer industry.
   When I arrived on campus, Fox's team had barely gotten a
four-processor prototype working. But since nobody knew how
to program the radically new machines, my first job was helping
figure out how to use them to solve problems that had previous-
ly been solved sequentially. My role was "speed hacking"-trying
to figure out clever ways to get more performance for a particu-
lar problem-something I had done a lot of at Princeton. One of
the things we discovered early on was that the hypercube com-
puters were ideal for computing a set of mathematical problems
known as fast-Fourier transformations, which are used in signal
                       THE REAL WORLD                           65
processing and have practical applications for everything from
hunting enemy submarines, to recognizing human speech, to
compressing data.
  I worked with Fox full-time for the summer, but after the
school year started, my interests moved in other directions, and I
quickly drifted away from the project. One factor that led me
away was a competing job offer from NASA's Jet Propulsion
Laboratory, just up the hill from Caltech, The JPL engineers
offered me the opportunity to work in communications systems
research, an esoteric area responsible for much of the work
involved in building radio bridges to deep-space probes like
Pioneer and Voyager that had been sent to other planets. Some of
the best communications people in the world were at the lab dur-
ing this period, and they were looking for students who were
adept at stepping in and working on projects in which there were
no easy guidelines to follow. My Unix and computing experience
turned out to be a much sought-after commodity. The JPL group
was trying to build a design system on a Unix machine to help
them do gallium-arsenide integrated circuit design, and I soon
became the resident Unix hacker and troubleshooter.
  At Caltech there is a long hacker's tradition of cleverly manip-
ulating the system. The criteria were that any hack had to be
done in good style; it should be clever, amusing, and not a copy
of something done before; and most importantly, it should not
be destructive or harmful.
   I participated in my share of pranks. For example, I had done'
some rock climbing before I came to Caltech, and once I arrived
at school I discovered that the fact the campus was in the middle
of a city didn't stop the school's climbers. Scaling buildings is
called buildering, and there is even a climber's guide to the archi-
tecture of Caltech. Rappeling off buildings at 2 A.M. was one of
our favorite sports for avoiding homework. Of course, campus
security hated to have people climbing around on their buildings,
and so throughout my undergraduate years a running game was
waged between the climbers, who were trying to avoid detection,
and the guards, who were trying to stop them.
   One evening a friend and I decided we would give the security
guards something to do. Since there was nothing wrong with
simply walking around wearing climbing gear, the two of us
66                         BREAK-IN
slung ropes and climbing hardware over our shoulders, and pro-
ceeded to amble leisurely around campus, stopping in front of a
few popular climbing routes and many improbable ones. Before
long we had collected a following of a handful of guards, who
stood and eyed us at a distance with their radios squawking, and
continued to follow us around on our tour of great climbing
spots, until an hour later we each turned and went home in
opposite directions.
   On the academic front, I began the school year trying to
behave like a normal student, hoping that college would be dif-
ferent from high school. But after the first few weeks I realized
that it was much the same experience, and I found I was focus-
ing on what interested me, and ignoring the fact that the school
expected me to jump the mandatory hurdles along the way.
   There were two classes, however, into which I threw myself
with real enthusiasm. One of them was a course taught by Ron
Drever, a general relativity experimentalist, who is well known for
his work with gravity wave detectors. The problems he's best at
solving involve detecting remarkably minute effects in an envi-
ronment cluttered with larger, more powerful forces. The class,
half juniors and seniors, the rest graduate students, except for me,
basically focused on how you could measure incredibly small
gravitational effects by cleverly setting up your experiments.
Much of our time was spent reviewing experiments in theoretical
relativity, looking at both effects that had been postulated but not
yet measured, and measurements that have actually been accom-
   One notable thing about the course was that it had no
midterm, and the entire grade rested upon designing a labora-
tory experiment to measure one of the as yet unmeasured
effects predicted by general relativity. I came up with an idea
for measuring a phenomenon called gravitational frame drag,
using a tool called a laser interferometer in a novel way. We all
turned in our papers, and Drever started off one of his last lec-
tures by announcing, ''I'm very disappointed in you. I only got
one original idea out of all the papers, and that was from a
   The other course that had a huge impact on me was a gradu-
ate course taught by Richard Feynman; Carver Mead, the father
                       THE REAL WORLD                           67
ofVLSI, or Very Large Scale Integrated Circuit design; and John
Hopfield on the physics of computation. Hopfield, who was one
of the inventors of neural networks, a computational model that
mimics biological systems, was one of my undergraduate advis-
ers, but it was Feynman's interest in the underlying foundations
of computation that particularly intrigued me. Feynman, one of
the world's leading theoretical physicists, hadn't taught during my
first quarter at Caltech because he was being treated for cancer,
but at the end of that quarter I introduced myself, and timidly
asked if I could take his upcoming course. He asked me a couple
of questions about my background, and then said it would be
fine for me to attend. I wound up taking the course both years I
was at Caltech.
   The seminar focused on the limitations of computing-quan-
tum limitations, communications limitations, coding theory
limitations, and thermodynamic limitations-and thus probed
the ultimate frontiers. Although I had explored parallel comput-
ing even in high school, through Feynman I began to see that
while modern computers processed information serially-one
instruction and one piece of data at a time-nature computes in
parallel. I also began to understand that serial computing actu-
ally corrupts the way you think as a scientist. Using serial com-
puters to explore a parallel world often masks the real simplicity
of nature.
   I spent the summer after my freshman year back in Princeton,
where I worked at the Institute for Advanced Study with Steven
Wolfram, a physicist who later developed Mathernatica, the pro-
gram that is now widely used in high schools and colleges. It suf-
ficed for summer work, but Wolfram was looking for a profes-
sional coder to help him develop software products, which was
not something that interested me. I write software, but I do it to
solve my own problems.
   When I returned to Caltech in the fall, I soon found that I was
growing bored by academic drudge work. I began taking more
upper division and graduate classes, dabbling in things, hoping to
find something in which I could completely immerse myself. In
the process, however, I quickly burned out. I was simply losing
interest in jumping through academic hoops for no apparent rea-
son. I did increasingly poorly in my required classes, and felt
68                         BREAK-IN
distracted. I was having more fun playing in the graduate classes
I was taking, and began to think about doing something else even
though I didn't have anything else in particular in mind.
  During my first year in Feynman's physics of computation I
had met Danny Hillis, the artificial intelligence researcher who
had recently founded the Thinking Machines Corporation in
Cambridge, Massachusetts. Feynman was a frequent visitor there,
along with a range of other scientists and engineers who were
intrigued by Danny's radical approach to building a massively
parallel computer. At the end of the school year, Hillis invited me
to Cambridge to work with Thinking Machines for the summer,
so Feynman and I represented the Caltech contingent at what
was essentially an MIT start-up company.
  The Thinking Machines Connection Machine computer was a
dramatic break with what had come before in high-performance
computing, a field that had been dominated by Cray Research.
Seymour Cray's machines were designed to use a small number of
very, very fast, and very expensive, processors. At Thinking
Machines, however, the notion was to divide problems so they
could be solved by more than 64,000 inexpensive processors
working simultaneously.
   I was able to work on a number of intriguing projects at the
company, but the one that was probably most useful was a sim-
ple hack for hooking together an array of small, cheap disks. One
of the biggest problems in supercomputing is getting the vast
amounts of data used in its calculations in and out of the
machine quickly enough. Using a group of inexpensive disks and
spreading the data among them, instead of relying on a single
fast, but expensive disk, was the perfect complement to the army
of inexpensive processors that were actually computing the data.
My contribution was to design a "self healing" array of disks-
that is, I figured out how data could be arranged across a number
of disks, so that if one failed, the data that had been on the sick
drive would automatically be regenerated on a spare.
   Hillis was a wonderful person with whom to work because he
was genuinely more interested in building machines that could
think than he was in being a successful businessman. He had
assembled a remarkable group of engineers and scientists, and
things often didn't take place in a predictable fashion.

                            THE REAL WORLD                            69
        One Sunday afternoon, for example, Danny and I discovered
     that we wanted something out of the office Coke machine, but
     that we were out of quarters. Danny went around the building
     hunting for the key, which he found, but we both decided that get-
     ting the key each time we wanted a soft drink wasn't an optimal
     answer to the problem. It seemed to us that we could design a more
     permanent solution: we would simply build an interface to the
     Coke machine so that you could control it from a computer con-
     nected to the Internet. It took us only about half an hour to get a
     serial interface working that would allow you to control the
     machine and credit yourself change from your computer's desktop.
     Our system went one step further than the classicCarnegie Mellon
     Coke machine hack, which was connected to the Net only in order
     to provide status information on how many cans were left in the
     machine and whether they were cold.
        I had a wonderful summer at Thinking Machines simply pur-
     suing whatever problem seemed interesting, and when I flew
     back to Caltech in the fall of 1984, the idea of being a student
     again was even less inviting than when I'd left in June.
        I'd had an offer to work with Steve Chen, Seymour Cray's com-
     puter architect, who would later leave to found Supercomputer
     Systems, Inc. I visited Chen at Cray Research, and toyed with the
     idea of going there, but locking myself into one company seemed
     in its own way as confining as school.
        At the same time, however, I'd gotten a call from a team of
     researchers who'd left Caltech for the Los Alamos National
     Laboratory in New Mexico to build a specialized parallel comput-
     er for physics research. Was I interested in coming to work there to
     join a newly created parallel computing project? It seemed strange
     that they would pursue an undergraduate when there were so many
     other people to choose from, but I realized that my breadth of
     experience in computing was valuable. I thought about my
     prospects for a while, and went looking for Feynman. I wanted his
     advice about whether I should stay on as a student.
        I found him one afternoon as he was walking across campus. I
     explained that my grades had put me in trouble with the admin-
     istration, and I didn't know if I wanted to stay anyway. He told
     me that if there was anything he could do to help my situation at
     Caltech, he would be glad to do it. I described the offer I had at
70                           BREAK-IN
Los Alamos, and asked his opinion. He wasn't going to make any
decisions for me, he protested, but I got the feeling he thought
I'd do better if! set out on my own. I decided it was time to leave
college for good.

I arrived at Los Alamos in late 1984 on a post-doctoral research
appointment, even though I had no high school or college
degree. At nineteen, I was the youngest staff member to come to
the Los Alamos theoretical division since Feynman had joined
the Manhattan Project in the 1940s. As the nation's oldest
nuclear weapons laboratory the lab was steeped in government
bureaucracy, and rife with bureaucrats, some of them my super-
visors. At the same time there was a "can do" spirit at the Labs
that I found refreshing, and there were pockets of intellectual
freedom in which it was possible to pursue interesting science.
  Although I had come to Los Alamos in the midst of the Reagan
Cold War buildup, within several years the nation's defense budget
would peak, and then begin to wind down, forcing the weapons
designers, many of them former physics prodigies themselves, to
justify their existencefor the first time in their careers. I, meantime,
took a YoungTurk's delight in knowing that I was siphoning fund-
ing away from the weapons budget into the far more intellectually
interesting area of basic physics research. Instead of puzzling over
problems like how to blow up the enemy more efficiently, I worked
with a group that spent its time exploring the very fundamentals of
computation, which had only theoretical ties to weaponry, and
thus put us out of the lab's mainstream.
   Although my original mandate at Los Alamos was to help
design a new kind of parallel supercomputer, I ended up becom-
ing part of the Theoretical Division's scientific visualization and
simulation team, led by a brilliant physicist named Brosl
Hasslacher, who was twenty-four years older than me, and in
every way my mentor. It was Brosl who recruited me away from
Parallel Computer Design and back to physics, and we worked
together in a rich collaboration.
   Although Brosl had an international reputation as a physicist,
many of his superiors at the lab didn't appreciate the significance
of his work. One winter, our team was exiled to a virtual Siberia,
                       THE REAL WORLD                            7I
a trailer gulag outside the main lab building. It didn't bother us
that the funky trailer wasn't designed to support clusters of com-
puter workstations and that it needed to be jerry-rigged to ensure
adequate electrical power. But because it was essential to stay con-
nected to the outside world, we had to string a coaxial computer
cable to another trailer that was already hard-wired to the main
laboratory network and the Internet.
   It can snow a lot in Los Alamos, so to protect the cable from
accidental damage after it was buried in a storm, we placed Day-
Glo orange road markers along its length and alerted the mainte-
nance department to its existence. A lot of good that did. The
next day a Laboratory snowplow came through and neatly sev-
ered it. We strung new cable and called maintenance again, but
the next time it snowed, the plow put us out of business once
   Tougher measures were clearly in order. I had the idea of wrap-
ping the cable in Kevlar, the same rip-proof fabric used for bul-
letproof vests and for mooring submarines. I took a length of
Kevlar cord, anchored one end to a concrete pillar, wound it
around the network cable and then fastened the cord's other end
to the wall of the neighboring trailer. So there, I said to myself.
   It worked, but a little too well. The next time a plow ran over
our line, the Kevlar caught it, the way the arrester cable snags a
fighter plane landing on an aircraft carrier, and the plow yanked
the side off our neighbor's trailer. From then on, however, the
plow drivers were more cautious.
   In the summer of 1985 Brosl spent several weeks visiting the
theoretical physicist Uriel Frisch in the French countryside near
Nice. The two were collaborating on a fundamentally new
approach to computing they called lattice gas automata. In the
 1930s the mathematician Alan Turing had proposed a simple
sequential device for solving mathematical equations that
became known as the Turing Machine. The power of the Turing
Machine is that it can simulate every other computational
scheme, and as a result it has become the standard tool for think-
ing about computation.
   However, both Brosl and Frisch were physicsts rather than
mathematicians, and they came upon a new parallel computa-
tional model drawn from a physicist's world view. They realized
72                          BREAK-IN
it was possible to model the flow of fluids in an entirely different
way than had been done before, and they began to think about
the design of computers that would be needed to simulate such a
model. The obvious payoff is that by computing in parallel it is
possible to achieve orders-of-magnitude increases in speed. In
their model, instead of computing a complex formula sequen-
tially, the flow of a fluid is· simulated by a system that is made up
of many simple components that interact locally. In other words,
an algorithm, or recipe, for sequentially computing a problem is
replaced by many independent agents that are called cellular
   Traditionally, for example, the flow of fluids has been described
by a complex equation known as Navier-Stokes. Now Brosl and
Frisch proposed the idea of a hexagonal array at each point of
which particles could be represented colliding, and in motion. A
set ofsimple collision rules for each point on the array are enough
to model everything that required a complex equation, and it can
simulate fluid flow in two or three dimensions.
   Frisch and Brosl were good friends and they both realized they
were on the edge of a dramatic advance, but Frisch was a
Frenchman first, and a quite nationalistic one at that. After sev-
eral days Brosl realized that every afternoon his friend was going
off by himself, and speaking by telephone to a team of four or five
programmers in Paris. He was trying to perform an end-run, and
give the French a headstart on being the first to implement a
working version of the lattice gas automata model!
   Brosl decided that two people could play this game, and so
one evening he phoned me back at Los Alamos and described
his and Frisch's basic theoretical model in detail. I proposed
some minor changes and told him that I thought I could imple-
ment it quickly. The machine I had to work on was called a
Celerity, a Unix scientific workstation with a 1280-by-l024-
pixel high-resolution display. I worked at coding for a couple of
days to implement Brosl's theory in a program that would
graphically display the flow of a fluid as it emerged from the
tens of millions of small particle collisions. Because I was just
modeling a small set of local rules about the behavior of the par-
ticles, the software was dramatically simpler than existing
                        THE REAL WORLD                           73
versions. The essential elements of the simulation could be
described in several dozen lines of code, and it was far less com-
plex than the several hundreds of lines of code normally
required to perform two- and three-dimensional hydrodynamic
  When Brosl came back from France a week later I had some-
thing to show him on the display, and it was almost there, but
something was not quite right. He came up with a few changes,
and then went home while I kept tinkering. About midnight I
phoned him.
   "Brosl, you better come up here and look," I said. "There's
something weird happening on the screen."
   On the computer monitor a thin line representing a plate
inserted to disturb the flow of the fluid as it passed around it was
surrounded by a halo of slowly changing colors. Brosl instantly
recognized that we had done it; the image gradually transformed
as millions of particle collisions were computed, and vortices
were clearly emerging. We left the image frozen on the screen,
and the next morning when we came back to the Lab the room
was crowded with hydrodynamicists who were startled to see
that we were computing something hundreds of times faster
than traditional sequential algorithms.
   Brosl's theory didn't gain instant acceptance, however. Entire
academic empires had been constructed on the old sequential
models, and the publication of his paper on lattice gas automata
in August 1985 created a nasty spat in the scientific community.
Some scientists initially attempted to challenge the accuracy of
the technique, but we were soon able to confirm our results. It
was remarkable proof that controversial parallel computing tech-
niques could provide dramatic speedups over existing approach-
   Despite his intellectual triumph, Brosl's work still remained
outside the mainstream, and in mid-1988 we decided to get away
from the politics and infighting of the weapons lab. We moved to
San Diego to set up a remote site of the lab's Theoretical
Division. With the end of the Cold War the weapons labs were
already beginning their decline, and as funding dried up, the
bureaucracy was becoming more and more imprisoning. Roger
 74                        BREAK-IN
  Dashen, who was a physicist I knew well, was trying to turn the
  University of San Diego physics department into a lively and
  eclectic place, and he offered me a research position there. One
. evening that summer Brosl and I finished loading up an 18-
  wheeler semitrailer with our computer gear and headed west,
  through the cool desert night.
On the morning after Andrew and I arrived in San Diego fol-
lowing the break-in, Room 408 of the Supercomputer Center
became our war room.
   A large room on the top floor, it once offered a view of the
ocean, which had been recently blocked by the new School of
International Mfairs building, and further obscured by a pair of
monitors, a camera, and other videoconferencing aparatus par-
tially covering the windows along that side of the room. But oth-
erwise the room's various features-a long conference table,
whiteboards on its walls, and network connections for our
portable computers-were perfect for our purposes.
   Around midday our impromptu force began to assemble.
Even though it was Christmas week and school was out, there
were always some researchers, students, post-docs, and even the
occasional administrator and technician around the Center, and
we were able to assemble an instant detective team for our effort
to recreate the break-in. To add a little incentive we had lunch
ordered from the Thai House, one of our regular spots, about
ten kilometers from the campus. I had decided that if Sid was
going to spring for expense money we ought to do something
truly useful with it-like feeding people.
   I had managed to round up with several phone calls an eclec-
tic squad of people who were willing to volunteer some of their
time. Because it was short notice it was something of a rag-tag
76                          BREAK-IN
army, and while some of the members of the group took a spe-
cific task, others were there just to lend moral support, or out of
curiosity. Rama Ramachandran was a former UCSD undergrad-
uate who was now in business school at the University of Chicago
and was visiting for the holidays. I was still puzzled by the strange
syntax error message from the X-NeWS PostScript interpreter
that had been on Ariel's console display the previous day, and
since Rama was a real PostScript wizard, we immediately set him
to work examining the interpreter to see if it had been exploited
to gain entry.
  Among those joining us were John Moreland, a scientific visu-
alization programmer; and Henry Ptasinski, a grad student in
electrical and computer engineering at UCSD. At the time,
Henry was also one of the system administrators at CERFnet, an
Internet service provider that is closely affiliated with the Center.
   And then there was a manager of network security at the
Center. Tom and I usually get along well as long as we don't have
to work together, but he likes to get involved in everything, and
sometimes I refer to him as a roving speed bump. I may seem
cruel in this regard, but I've never been able to figure out how to
be tolerant dealing with the people who are responsible for main-
taining the needless red tape and regulations that seem to make
any large organization run. Julia says that my favorite word is
"bozo," and argues that I should work on being a little more
diplomatic. I tried to do so now, when he showed up to see if he
could do anything useful. I gave him a chunk of code called the
"remote procedure call ToolTalk database server daemon," or
rpc.ttdbserverd-which facilitates the communication of some
programs across a network of computers-and asked him if he
would go off and analyze it for vulnerabilities we didn't know
about. We had become suspicious that it might have played a part
in the break-in, because one of our activity log files had recorded
an unusual access to it one Christmas night.
   When he left the room, one of the graduate students turned to
me and said, "Why did you give him something like that to do?
You're always complaining about him."
   Andrew and I looked at each other, and I replied, "Basically to
keep him out of trouble and out of our hair."
   "You know he's not going to stay out of your hair," the student
                         FORENSIC DATA                            77
   "So?" I said. "This will keep him busy for a while, at least."
   We were still doing data collection at this point, and things
were looking pretty bleak, which made me even grumpier. I had
already taken data off Rimmon and Astarte, my machines at
home. At this point we could tell that Osiris had been tampered
with, but not how. On each of these computers we had checked
to see if any files had been altered and if any questionable pro-
grams had been left behind. When we didn't see any immediate-
ly I began to worry even more, for it suggested that our interlop-
er knew some other way to get in, and that he thought he'd be
able to return undetected. I couldn't consider going back on-line
until I assessedwhat our risk of another break-in was. Everybody
went off to his individual task, leaving Andrew and me working
at the portable computers we had set up.
   Progress turned out to be sporadic. Thanks to Ariel's age, get-
ting useful data out of the computer was frustrating and took us
much of the day. Most of the components in a modern comput-
er are connected by a bundle of wires known as a data bus.
Microprocessors, computer memory, disk drives, graphics dis-
plays, and assorted peripherals all plug into this main highway or
backplane, which is really just a set of parallel wires that enables
information to shuttle back and forth with incredible rapidity.
Ariel was so old that it used a bus called VME which was origi-
nally invented for minicomputers back in the 1980s. Its disk dri-
ves were also based on an outdated standard so there was no way
that we could hook Ariel's disks directly to my portable comput-
ers, which used more modern SCSI disk drives. As a result, we
first had to copy all the necessary data off Ariel before we could
work with it safely.
   The two of us went rummaging around the building trying to
borrow some extra disks to accommodate this huge amount of
data until finally, late in the afternoon, we managed to scrounge
up a two-gigabyte drive from the folks at CERFnet. (To under-
stand just how much information two gigabytes represents, think
of one of these drives as having the capacity to comfortably store
the entire Encyclopaedia Britannica, both pictures and text-all in
the palm of your hand.) More hours were taken trying to figure
out how to move all the data so that it would be organized exact-
ly as it was stored on Ariel.
   It wasn't until 10 P.M. that I had the data from Ariel transferred

       78                          BREAK-IN
       to the hard drive and ready to be examined on my RDI portable.
       By that time almost everyone else had left the Center, and
       Andrew and I felt the need for a dinner break. We took the ele-
       vator downstairs and drove off campus to Rubio's, an inexpensive
       fish-taco restaurant, where neither of us was particularly happy to
       be. "Look, we're on someone else's expense account and we real-
       ly should be getting better meals than this," I told Andrew. "I
       don't want to present Sid with a dinner bill for $4.95." But after
       ten o'clock in that part of San Diego, there aren't too many
       choices. We ate quickly, eager to return to the Center and see
       what Ariel's data would tell us.
          Back in Room 408, it took me about an hour to run the foren-
       sic-analysis programs that, as with Rimmon and Astarte, at home,
       revealed which of Ariel's files had been accessed, modified, or cor-
       rupted. For the first time I knew what had actually been stolen
       from Ariel: virtually everything in my directories. Much of it was
       valuable to me and my work, including tens of thousands of my
       e-mail messages, source code for programs I had written, and
       sensitive proprietary data. There were no conclusions to be
       drawn from this bill of particulars, however, because the thief or
       thieves were so indiscriminate that they had also spent hours
       copying programs that are freely available elsewhere on the Net,
       including various tools that I myself had downloaded from the
       Free Software Foundation.
          Our analysis of Ariel's data did turn up one piece of news: the
       intruders had been stealing files only two hours before Andrew
       discovered the break-in. So now we had a fairly complete picture
       of what had happened, and some indication of when. But we still
       couldn't answer the question that, to me, was the more important
       one: How did they do it?
          I knew that Osiris, the machine at the head of my bed, had
       been accessed before Ariel, at my SDSC office, but I didn't know
       how they had gotten to either machine, or if they had used one
       to get to the other. And then there was that XNeWS PostScript
       interpreter error message on Ariel, which seemed to indicate an
       attempted probe from Colorado SuperNet. Significant-or a
       false clue? If our attacker really knew what he was doing, disin-
       formation was a possibility we had to be aware of.
          There were also other bits and pieces that made up a puzzle I
                             FORENSIC DATA
    still couldn't assemble. One of them was a mysterious program,
    Tap, that I had seen when I peered into Osiris's memory the day

    before. It was a transient program that someone had created and
    placed in my computer's memory for a specific task. When the
    computer was turned off or rebooted it would vanish forever.
    And what about the ghost of the file oki.tar.Z, whose creation
    suggested that someone was after cellular telephone software,
    despite the otherwise indiscrimate nature of the looting?
       There was another crucial discovery from looking at Ariel's
    data; the intruder had tried to overwrite our packet logs, the
    detailed records we keep of various packets of data that had been
    sent to or from our machines over the Internet. The erased log
    files revealed that in trying to overwrite them the intruder hadn't
    completely covered over the original file. It was as if he had tried
    to hide his footprints in the sand by throwing buckets of more
    sand on top of them. But here and there, heels and toes and even
    a whole foot were still visible. It appeared that we had our first
    clues. We might not have the entire getaway path, but at least
    we'd know in which direction to start tracking.
       In fact, while none of the puzzle pieces fit yet, the packet log
    gave us a potential way to start putting our clues together. The
    intruder's alteration of the separate activity log had first alerted
    Andrew to the break-in. Now the clumsily overwritten packet
    log file could potentially help us recreate it, thanks to the tech-
    nology used for routing chunks of data-packets-over the
       This technology is called "Packet switching," and like the
    Internet itself it is the direct outgrowth of an idea conceived in
    the early 1960s by a Rand Corporation researcher named Paul
    Baran. In those days, at the height of the Cold War, the military
    was obsessed with the problem of surviving a nuclear war, and so
    one of the assignments they handed to their think tanks was
    inventing a communication system that would continue to oper-
    ate even if some of its nodes were destroyed.
       Baran hatched the idea of a computer network that could auto-
    matically reroute traffic. The technique, packet switching,
    involved breaking each message up into lots and lots of small
    packets. Each packet contained only a small portion of the mes-
    sage, accompanied by a "packet header" that contained enough
80                          BREAK-IN
information so that at each waypoint in its passage through the
network, each of these little data parcels could be rerouted if nec-
essary and still arrive safely at its final destination. The routing
computers were given enough intelligence so that even if the
packets took different routes and arrived out of order, or were
even lost, it was possible to reassemble the message in the correct
order and request that lost packets be resent.
   Baran's was a brilliant concept, and in the late 1960s the
Pentagon's Advanced Projects Research Agency, or ARPA, fund-
ed an experimental project to develop such a network. The first
"Watson, come here. I need your help" message was sent between
Stanford Research Institute in Menlo Park, now SRI
International, and a group of computer researchers at UCLA in
1970. Since then, things have gotten a little out of hand: from
the original two ARPAnet sites, the Internet network had
expanded to more than 6.6 million machines and was still grow-
ing at a geometric rate.
   But while the multitude of machines and users are overbur-
dening the Internet in various ways, and providing cover for peo-
ple intent on mischief, each of the billions of packets flowing
through the network still has that informative header, telling not
only where the packet is going, but where it supposedly came
from. And because I knew that a packet filter can duly note all
this information, I was hoping that the Ariel's packet logs might
eventually help recreate the intruder's actions.
   But there was a complication: although the intruder had
failed to overwrite the packet-log file, his effort to erase it was
going to make reading the data difficult. The way computer
data is stored on a hard drive is similar to the way a library orga-
nizes its holdings. What you really want in a library is to be able
to go and ask the librarian for a particular book, and then be
handed the book-you don't care where it's actually stored. In
the same way, information about the files that you create on a
computer is all stored in one place on the hard disk-think of
it as a library card catalogue-but the information itself is kept
somewhere else, usually scattered in little blocks all over the
disk's surface.
   Like librarians, computer operating systems take care of the
tedious grunt work of storing and hunting for the information.
                          FORENSIC DATA                            8I
When the operating system erases a file, what it actually does is
erase the pointers to the information, the file card in the catalog,
rather than the information itself, which remains until all available
space on the hard disk is filled up and the erased data eventually
gets written over by newly stored data. (Trying to prevent such
write-overs was one of the reasons I had Ariel and the other
machines halted as soon as possible after learning of the break-in.)
   So even though the packet file had been erased, it was actually
possible that its data could still be reconstructed from the disk-
it was just that the task was a little bit like the one the king's men
faced with Humpty Dumpty. As a first step in how we would
proceed, Andrew said, "I think I can write a program that will
find the spot in the file where the corruption ends and then look
for the place the real data begin." This would be a useful start,
but it wouldn't necessarily let us find all the disparate pieces of
the data we were looking for-as this would only turn up data
written to the file after 'it had been tampered with.
   It occurred to me that there might be a better, less obvious way
to find the same data. As a physicist I think a lot about concepts
like entropy and chaos, and I have spent a lot of time building
tools that look for patterns that might not be otherwise apparent.
A body of data may appear to be noise, but in fact may have a
hidden structure. The challenge is to extract that structure, which
may exist in a clear form, or in some form that requires the right
filter to see.
   In a way I was facing the same problem that a Roman cryp-
tographer might have confronted in attempting to break an
ancient coding scheme known as a Caesar cipher. In this proce-
dure, military messages were written on the surfaces of parch-
ment that had been wrapped around a cylinder or cone. The only
way to decode the message was to find an object of the same size
and shape and to keep wrapping the paper around it until the
writing lined up.
   Likewise, I needed to find a pattern in the tiny pieces of data
scattered across the surface of our disks. Like all computer data,
it was in the form of binary code-strings of 1's and O's that can
represent digits, letters, and other types of information. Each
piece of string was a link in a chain of information; the problem
was to discover the pattern by which these individual links had
82                         BREAK· IN
been scattered, so we could find them and reassemble the chain .:
I had put off doing this, because it seemed like a long shot. But
we hadn't learned enough from our other analysis thus far, so it
now seemed like the necessary next step.
   "Let's see who can do this first," I told Andrew at about 1:30
A.M. We agreed that Andrew would write his conventional pro-
gram for retrieving the packet information, while I would write
one to look for patterns on the disk and then attempt to reassem-
ble them into something that resembled the original file. We set-
tled in front of our workstations at a conference table, facing each
other, Andrew tapping away at his RDI portable, and I at my
newer version of the same computer, a machine in which I was
growing increasingly confident, despite its being an unproven
   I wrote a program I called Hunt to search the disk that I ran for
the first time at about 2:45 in the morning, and a second program
called Catch, which was designed to organize what Hunt found. I
actually won the race by a nose, with my programs finishing their
tasks at nearly 4 A.M., just ahead ofAndrew's. In the end, we both
succeeded in retrieving data, and Andrew's partial file was a useful
check of my catch of relevant data: 14 million bytes that had been
scattered among nearly 2 billion others and which might now
finally enable us to recreate our intruder's actions.
   Savoring the moment, I sat back and casually scanned through
our reconstituted packet-log file. With this information, we
might have a chance to replay his actual keystrokes, much like
rewinding a videotape to watch a television show. We now had a
chance of putting the puzzle back together. It was the first time
that I had been able to feel good in three days.
   My intruder had assumed that by overwriting the data he
would make it disappear. He should have known better. "He's
probably an MS-DOS user," I grumbled. If he was trying to be
invisible he shouldn't have been sloppy. I began to wonder how
good he really was. One of the standard strategies in the com-
puter underground is to share "cookbook" recipes for attacks, and
then use these step-by-step programs against targets over the
Internet. There's a lot of that: someone steals a hardware or soft-
ware company's source code, or stumbles upon garden-variety
computer-security software like the type that had been among
                         FORENSIC DATA                            83
my stolen files, or studies computer science journals and figures
out a way to break into a system. If he succeeds he passes word
around the network to his friends, or posts the how-to details on
any of the various underground bulletin boards that function as
rogues' hangouts on the Internet. Maybe our intruder was just an
anklebiter who had learned how to read technical manuals or
bulletin board postings, and hadn't realized that covering your
trail in the digital world isn't always as easy as it seems.
   However promising our leads, I was on my third night of little
sleep, so we agreed to quit the search. I drove home, and as my
Acura glided through the deserted streets I took satisfaction in
knowing that even if we didn't find enough information in the
packet log file to start chasing our intruder in earnest, at the very
least we should be able to figure out how he broke in, and so find
ways to improve our locks. When I got home the first hint of
sunrise was filtering through my bedroom, but despite my phys-
ical exhaustion, sleep eluded me. I sat crosslegged on my futon
bed in front of Osiris while looking through our systems for
other clues. I played with rpc.ttdbserverd. Why had it been left
running on the night of the break-in? Did the attacker have some
clever program that spanned an entire computer network to
break security? It nagged at me, but after another hour of fruit-
less searching, it looked like a dead end.

Late that morning, as my car climbed the hill to SDSC, instead
of feeling dread, I was anticipating the challenge ahead. Andrew
had already arrived in Room 408, as had most of the others, and
lunch had been brought in again. While we were eating, the
phone rang in the conference room.
   It was Mike Bowen, someone I knew at CERFnet. Mike, a
technical support person and a wizard with a type of digital tech-
nology known as ISDN telephone technology, also kept his ear to
the ground for rumblings from the computer underworld. I had
talked to him the day before on the off chance he might have
heard something about our break-in. He had told me he knew a
guy named Justin Petersen, whom I'd heard of, who was in prison
in Los Angeles for credit-card fraud and was trying to strike a deal
with Federal prosecutors. Petersen had been trying to persuade
84                          BREAK-IN
Mike to use his contacts in the computer-security community to
see if someone could persuade the Feds to listen to his alibi-that
he had been set up by Kevin Mitnick while he was trying to help
the FBI find Mitnick. Maybe I'd want to chat with Petersen?
"Sure," I had told Mike. "Why not?"
   Now Mike was calling back, saying he had set things up.
Because Petersen was in prison he was allowed phone calls only
from a restricted list of people. For him to speak to anyone else,
a person from this list would make the phone call and then three-
way the other party into the conversation. That was about to hap-
pen, Mike said, and told me to stand by. He hung up.
   A few minutes later, the phone rang again. "Hi, I'm here with
the person who you were expecting to hear from," said a voice I
didn't recognize.
   "Who are you?" I asked.
   "Why don't you call me Eric," another voice on the line
responded. Petersen had decided to use one of his many aliases,
although another, Agent Steal, was the best known.
   Justin Tanner Petersen was a strange character. A Southern
California native, he had originally been arrested in Dallas in
1991 and charged with credit-card fraud and other computer
crimes. In a plea bargain, he arranged with the Secret Service and
FBI to be released to work under Federal supervision to help
agents track down computer criminals, while other charges
against him played out in the California courts. Petersen had sup-
posedly put the FBI on the trail of Kevin Mitnick in 1992, forc-
ing Mitnick to go on the lam. And he had also helped law-
enforcement officials assemble evidence against Kevin Poulsen, a
Silicon Valley programmer who had been arrested in 1991 and
who finally pleaded guilty in June 1994 to electronically taking
over a Pacific Bell central office telephone switch to rig contests
at two Los Angeles radio stations, from which he won two
Porsches, more than $20,000 in cash, and at least two trips to
Hawaii. (If you are in control of the telephone company's central
office switch, you can be lucky caller number ninety-five when-
ever you want.) Meanwhile, the FBI had an extensive dossier
against Poulsen for other computer and telecommunications
activities, such as listening to the phone calls of his former girl-
friend, tapping into conversations of telephone company securi-
                         FORENSIC DATA                            85
ty officials investigating him, and even monitoring the electronic
communications of FBI agents who were tailing Imelda Marcos's
daughter in Woodside, California.
   But while working for the FBI, Petersen had allegedly returned
to computer crime himself. In an October 1993 courthouse
meeting with a prosecutor for the Los Angeles district attorney's
office, he admitted to credit-card fraud. Then, in the middle of
the meeting, he told his attorney he needed to take a break,
walked from the room, and fled. He lived on the run until being
recaptured in August 1994, and now, more than four months
later, he was about to be sentenced and was hoping that I could
help him make a deal for a lenient sentence in exchange for his
helping us catch Kevin Mitnick. Although he'd been trying to
negotiate with the Justice Department, his prospects seemed
grim. Because I knew Scott Charney, the Justice Department's
chief computer crime prosecutor, Petersen seemed to hold out
some hope I could help him in making a deal.
   He believed it was Mitnick who had turned him in to the Feds
in his most recent arrest, and he didn't sound happy about it. He
had a flat Southern California accent and I had the distinct sense
that he wasn't being very sincere.
   "I don't even know if it was Kevin Mitnick who broke in to my
computers," I said.
   "It sure sounds like Kevin's M.O.," Petersen responded.
   I was skeptical. There were potentially thousands of people
who could have gone after my machines. "What would you need
to find him?" I asked. "I understand that you're in a difficult sit-
uation, given that it seems you screwed over the Feds at least
once. "
   He was vague. "I know things that I obviously don't want to say
over this phone line," he replied.
   Petersen began proposing that we meet in person, and then he
would be able to tell me more. He said he thought he was very
close to catching Mitnick, that it would take maybe a month. He
proposed expense money. He was hard to read, and I kept trying
to figure out whether he actually had something or not.
   After we talked for about forty-five minutes I finally said, "If I
get a chance I'll mention this to to law enforcement people, but
 I don't think it will lead anywhere." I added that if I was in Los
86                          BREAK-IN
Angeles I would come visit him in prison. We hung up, and I
called Mike Bowen and said, "Should I believe any of this?"
   "I don't know, maybe," Mike answered. Perhaps Mitnick real-
ly had set up Petersen. "But there's another possibility," he con-
tinued. "Maybe Justin is worried that Kevin Mitnick has the
goods on him, enough to put him away for a long time."
   I decided that for now, at least, I might be better off analyzing
Ariel's data than relying on people like Justin Petersen.

Shortly after 5 P.M. Andrew and I were ready to begin recreating
a second-by-second chronicle of the events of the break-in on one
of the large whiteboards along the wall. We had attracted a small
audience of curiosity seekers who had heard about our project,
including Jay Dombrowski, the Center's manager responsible for
networks and communications.
   We were at the crux of our investigation, and doing it so pub-
licly was something of a risk: it wouldn't look very impressive if
we came up empty-handed. But the opportunity for us all to
learn something outweighed the risk of embarrassment.
   During the afternoon I'd massaged the reassembled packet log
file with a program I'd written called Cook that threw out any-
thing extraneous. I'd also assembled the various forensics data
we'd collected-primarily, our records of the files that had been
accessed-and merged it into one file, organized chronologically,
that gave us a single timeline of all events. The packet log was
already organized chronologically. Everything we had done over
the last several days had been preparation for this: we would now
systematically compare the packet logs, which would show us
exactly what the attacker typed or transmitted, with the forensics
data, which would reveal the consequences of each of these
   Andrew stood at the whiteboard with a magic marker, and I sat
in front of my RD1 workstation. I began calling out each event
as I extracted it from the lists we had compiled and merged.
   I began with the afternoon of Christmas day-shortly after I
had wandered by the computer in the entryway at Toad Hall and
thought about checking mail on my network.
                         FORENSIC DATA                           87
  "14:09:32," I called out. From our reconstructed packet data
we saw that Ariel received this command over the Internet, an
exploratory probe:

       finger -1 @arie1.sdsc.edu

   Finger is a standard Unix utility that displays information
about each logged-in user, and Ariel responded by giving basic
information, telling the prober that there were current connec-
tions to Astarte, Rimmon, and Osiris, and that my computer had
been unattended for several days. Over the next three minutes in
our computer timeline, I called out six more probes, each target-
ing a different aspect of my network.
   "14:11:49," I read. "Hey, they've run a remote procedure call
on Osiris."
   Andrew walked around the table and studied the screen of my
portable. He was an expert on remote procedure calls, RPC's, an
operating system function that lets programs ask a remote com-
puter to do something. The result he was studying was displayed
in hexadecimal format, the base-16 numbering system that good
hackers learn to read as a second language. "That's a showmount
-e to show exported file systems," he said. In other words, it was a
command that allowed the person executing it to determine
which hard disks were shared by the other computers in my net-
work. Someone was trying to build what is known as a trust
model of my network, to see which computers had special rela-
tionships, with few security barriers between them. It was an
attempt to see which computers in my network "trusted" each
other, as Osiris and Rimonn did, for example.
   I looked more closely at the probes and made a striking dis-
covery: They all came from toad. com.
   "This is very strange," I said to Andrew. "I was at Toad Hall
when these probes were made, not ten meters away from the
machine they came from." I could see that the RPC had come
from source pon 721 at toad.com, which meant it had been
issued by someone who was root on Toad. I knew there had been
 no one else physically present at Toad Hall at the time but Julia
88                          BREAK-IN
and myself, and I realized that the attack could have been staged
from anywhere on the Internet. Still, I couldn't help wondering
whether the intruder was someone I knew.
   I was puzzled, but there was nothing to do but plunge ahead.
   Six minutes later in the data stream we saw evidence of some-
one attempting to initiate an Internet connection-a request
called SYN, for synchronize.
   "14:18:22," I said. "I see a remote log-in connection from to Rimmon ... wait, there are a whole bunch more!"
I was struck by this. Normally, a SYN request should have initiat-
ed a single computer handshaking sequence-the brief greeting
and interrogation between two machines before they agree to com-
municate over the Internet. This requires the pair of computers cre-
ating and exchanging a sequence of one-time-use numbers to
ensure that this conversation isn't confused with any other simul-
taneous coversations that either computer may also be having.
   But in this case, it was as if this remote machine was saying
"hello," "hello," "hello," "hello," in rapid succession, without lis-
tening to Rimmon's reply. Why would that happen?
   I stopped and tried to find where these rapid-fire SYNs had
come from. The digits were the remote computer's
Internet address, and it took several queries to various Internet
databases, but I finally got an answer: currently, there was no
such computer. The messages to Rimmon had appeared to come
from a network in Switzerland:

       university of Berne (NET-UNIBE)
       Institute of Informatics and Applied
       Laenggassstrasse 51
       CH-3012 Berne

       Netname: UNIBE

       Buetikofer, Fritz (FB61) btkfr@ID.UNIBE.CH
       +41 31 65 3843
       Domain System inverse mapping provided by:
                          FORENSIC DATA                            89
  That network existed, as indicated by the first five digits: 130.92:
But it appeared that the computer that had tried to connect to
Rimmon, the machine designated by the complete address,, didn't answer, or didn't exist, at least not now. The
computer could have been turned off since the attack, I supposed,
and so would not be showing up in the database. Or there was
another possibility:The address could have been a forgery.
   I resumed the chronology: "14:18:25." It was just three sec-
onds later, in our data timeline. And now there was another SYN,
this time to Osiris from a computer called apollo.it.luc.edu. I
queried the Internet database again and found out that luc.edu
was Loyola University in Chicago. As had happened to Rimmon
from the mystery Swiss machine, Osiris was now receiving a
series of remote log-in connection requests from the Loyola
   "This is very weird," I muttered. What was going on? Osiris
was receiving a series of SYNs, each with a sequence number to
initiate the handshake. But when Osiris sent back its acknowl-
edgment-SYN-ACK, which included a second sequence
number-the Loyola machine was not taking the normal next
step. Instead of replying with a third sequence number, the
Loyola computer would start the process over by issuing the
command RST, for reset. This happened twenty times in rapid
succession. Why?
   I continued to hunt through the data, and then I saw some-
thing that at first didn't make any sense at all. All of the packet
data that we were analyzing were packets that had come in from
the Internet through Ariel, which was sitting in the wiring closet
here at the Supercomputer Center. But now, our records had
begun showing traffic that seemed to be flowing directly between
Osiris and Rimmon, inside my house. "Wait, I shouldn't be see-
ing these packets!" I said. "Why am I seeing local traffic between
Osiris and Rimmon?"
   But suddenly it came to me-the answer I had been pursuing
around the clock for the last three days. The remote computer
had seized on the fact that Osiris trusted Rimmon and had forged
a one-way connection to Osiris that appeared to be from
Rimmon, but was actually coming directly from our intruder.
   ''Ah, I understand," I said. The room fell quiet, as I looked at
Andrew. "So that's how they broke in."
    90                         BREAK·IN
       All those aborted handshakes now made sense. The attacker
    had needed to be able to predict the sequence number that Osiris
    was sending out with each SYN-ACK. A sequence number in this
    case was simply an authenticator, much like the number they give
    you while you're waiting in line at the delicatessen, so that when
    it's your turn to talk to the man behind the counter he and every-
    body else recognize your right to proceed. Our intruder planned
    to masquerade as Rimmon, a computer that Osiris trusted, and
    in order to do it successfully he needed to be able to send Osiris
~   back the sequence number, the delicatessen number, that it,
    Osiris, would expect Rimmon to present.
       And now I understood why the intruder had sent that first flur-
    ry of messages to Rimmon. They had filled up an input queue, in
    effect gagging Rimmon so that it couldn't respond when it came
    time to present its sequence number. Once Rimmon was bound
    and gagged, the attacker had then sent that series of twenty SYNs
    to Osiris, to learn the formula by which Osiris generated its
    sequence numbers-each was 128,000 larger than the previous
    one-and thus be ready to slip into Rimmon's place in the deli
    line and respond with the appropriate sequence number. The
    intruder then gave the sequence number Osiris was expecting,
    and used it to open a communications channel.
       Andrew had come over from the whiteboard and was watching
    my screen over my shoulder. Now that they had broken in, what
    did they do? Pretending to be Rimmon, the attacker at the Loyola
    computer had sent the following short message through the one-
    way channel: "echo + + >/.rhosts." That simple command caused
    Osiris itself to drop all of its defenses and made it possible for
    anyone to connect to it without a password. The intruder had
    convinced Osiris that it was opening up a digital conversation
    with its trusted file server in the next room, Rimmon.
        It was now almost six in the evening, and Andrew had returned
    to write up the sequence on the whiteboard. Jay Dombrowski,
    who was following some but not all of the chronology that we
    created, politely excused himself to go home for dinner.
        I thought about it for a minute and realized that the style of
    attack was a familiar one. With a deft sleight-of-hand the attack-
    er had made packets coming from outside our network appear to
    be coming from the safety of the inside. This was an "IP spoof-
    ing" attack, a type that had been described in theory in the com-
                         FORENSIC DATA                            9I
puter science literature but which, as far as I knew, had never
before been carried out as a hostile attack.
   The attack was based on a weakness in the set of technical com-
munications instructions for Internet traffic, known as the
Transmission Control Protocol/Internet Protocol (TCP/IP),
which had been developed in the late 1970s and early 1980s. IP
spoofing, manipulating the handshake sequence numbers to
impersonate another computer, was possible because the hand-
shake procedures, created in an era when nobody worried much
about Internet security, had been designed merely to clarify who
was who on the Internet, not verity,
   I was familiar with a technical paper on TCP/IP security prob-
lems, written in 1989 by Steve Bellovin, a computer security
researcher at Bell Labs, in which he had described how an IP-
spoofing attack might play out. But the potential to use IP-spoof-
ing to pose as a "trusted" computer had first been brought to the
research community's attention even earlier, in a 1984 paper that
a student named Robert Tappan Morris had written while a sum-
mer intern also at Bell Labs. On the final page of his report
Morris had given a cookbook description of how such an attack
would work. More than ten years later, his paper appeared
prophetic: "Bell Labs has a growing TCP lIP network connecting
machines with varying security needs; perhaps steps should be
taken to reduce their vulnerability to each other."
   As night fell, Room 408 was bathed in cold fluorescence as we
continued following the attack's digital trail. One of the things I'd
discovered on Tuesday was that on both Ariel and Osiris the
intruder had inserted a program directly into the computer's
operating system memory. Sun Microsystem's Unix has a stan-
dard feature that permits you to modify the heart of the operat-
ing system, while it is still running, to add new functions. These
programs are called "kernel modules" and they can be directly
placed in software "slots" in the operating system while the com-
puter is still running. Typically you might use one if you were
adding some new peripheral to your computer. The one on Ariel
just seemed to be garbage, but I had tried to take the one I found
in Osiris's memory apart and had not been able to immediately
 tell very much about what it was designed to do. It did have a
suggestive name however, Tap 2.01.
    At the time I'd wondered if it was a "sniffer" program, to enable
92                          BREAK-IN
the intruder to subsequently monitor the traffic over my net-
work, looking for things like passwords that might abet subse-
quent break-ins to my machines or the computers of other peo-
ple who communicate with me. But now in the traces of our
packet data I could see what had happened. After installing and
running a backdoor program on Osiris, the intruder returned
through his backdoor network port, a separate channel that hap-
pened to be one that our packet logs weren't monitoring, and so
we lost direct traces of his keystrokes. But during this blind spot
in our packet-log data, we could still follow his activities, the con-
sequences of his keystrokes, by consulting our forensics data for
this same time period.
   We could see that he had inserted a kernel-module program
called Tap into a slot in Osiris's operating system. Almost imme-
diately afterward we could see his activity jump from Osiris at my
home to Ariel at the Center. Although Osiris and Rimmon had a
trust relationship that made them vulnerable to IP spoofing,
Osiris and Ariel didn't. Initiating a communications session with
Ariel would have required a much more elaborate set of proce-
dures, including a password. He had needed another strategy, and
this is where Tap came in. As he could see with finger, my home
machine already had an open session to Ariel running on its
screen. It appeared that Tap had allowed him to literally hijack
this open window on Osiris's screen, and use it to control Ariel.
Tap was a program that gave my thief the power of a puppet mas-
ter, sending keystrokes through the portal just as if he had been
sitting on my bed.
   It was after midnight, and Andrew and I were once again the
only ones working late in Room 408. The inactive videoconfer-
ence terminals along the windows stared at me opaquely, and
suddenly I recalled how the blank window on Osiris's display had
puzzled me two days earlier. It was now obvious: my trespasser
had broken through that screen portal in much the same way a
burglar would jimmy open a window and then climb through it.
And once he was inside Ariel, he had been able to help himself to
my software and e-mail messages, and then carry them away to
who knew where on the Internet.
In the days immediately after I unraveled the IP-spoofing attack
my life didn't return to normal, for there was too much cleanup
and reconstruction work to be done. But I did find time for skat-
ing in the Southern California winter sunshine and talked with
Julia regularly by telephone, as we discussed the possibility of her
coming to visit me in San Diego. Much of my time was given
over to building a more secure router for my network that would
not only keep attackers out but would also keep detailed log files
and alert us quickly if we were attacked. Andrew also worked
long hours, deciphering the programs that had been left behind
by our data thieves, and together we spent several days mopping
up final details and trying to make certain that we understood
exactly how our security had been violated.
   I called Toad Hall to ask John Gilmore about the early probes
from toad. com. He had become increasingly uncomfortable
about my contact with Julia. It was a strained conversation. I
told him about the attacks on my computers and the initial
probes from his computer. He looked at the accounting logs his
computer kept, reporting there was no suspicious activity.
   "Youknow as well as I do that if somebody broke into toad they
could have doctored your log files to hide themselves," I said.
   Later, I spoke with Julia and we both decided the attack was
an amazing coincidence. We knew I hadn't been involved, but I
had been just upstairs from the computer, and we realized that
94                          BREAK·IN
in raising the issue we might be opening a can of worms. Was
somebody trying to frame me, or for that matter, to frame John?
Or was it something else entirely? We decided it would be best
not to make a point of it, as there were so many erroneous con-
clusions that people might jump to.
   While I was in San Diego both Andrew and I began working
on refining my network's defensive perimeter. For many people,
security in the Internet today is simply a matter of going out and
buying a system called a firewall, a black-box solution that just
restricts the type of data packets that can flow from the outside
world. I've never believed that merely building thicker castle walls
will offer a better defense, so instead we set traps and built alarms
into the network, making it easier for us to detect and respond to
future intrusions. Ariel was upgraded yet again, this time with the
installation of more modern disk drives. We also wrote software
that would protect against any future IP spoofing or similar
attacks. We wanted to be automatically notified if someone
attempted a spoofing attack in the future, and we began to mod-
ifY our network software so that it would be impossible to fool
our machines with a forged Internet address.
   Our new security router was designed to look at the address of
each packet that flowed from the Internet to my network. If it
found an address that appeared to be coming from the inside it
would say, in effect, "Wait a minute, this shouldn't happen," and
then not only reject the packet but also set off an alarm at the
same time.
   I scrounged around and salvaged a bunch of spare parts, and
configured the secure routing computer to sit on the network
between the outside world and my piece of the Supercomputer
Center network. It was built out of a Sun SPARCstation we
commandeered for the task, which took up residence in the
wiring closet next to Ariel. We actually gave it three names. If
you wanted to send packets to the outside world you sent them
into "chaos." If you were sending packets to my computers then
you leave them to "chance" the whole router we named the
   At the heart of our defense was a basic computer network tech-
nology called packet filtering. The ability to scrutinize and cap-
ture individual packets as they moved by on a wire first emerged
                   MY CHRISTMAS VACATION                           95
in the early 1980s, because network designers needed a diagnos-
tic tool to monitor and tune their systems. More recently, how-
ever, packet filtering has become a powerful tool with great
potential for abuse. Neither the first Ethernet local area network
nor the first Internet computer networks were built with privacy
or security in mind. They were simply research projects designed
to let computer scientists and engineers explore the idea of hook-
ing computers up within offices and between cities and states.
But between the late 1960s and today, computer networks
evolved from being research tools to the point that they've
become an integral part of the fabric of our society. Ethernet
works by broadcasting every packet along the entire wire.
Normally, computers on the network listen to the packet broad-
casts and simply take the packets addressed to them. The prob-
lem with Ethernet technology is that someone can take over a
computer on the network and simply scoop up all the packets,
whether they are addressed to him or not. This data is usually not
encrypted and it is a tremendous security loophole because sniff-
ing is a passive activity. There is no way to know for certain
whether packets bound for your computer are being illicitly
snatched up and scanned by someone else.
   As more and more data began to be transferred over computer
networks, somewhere along the way the bad guys began using
packet filters, or "sniffers" as they became known, to watch all the
traffic flowing over a network, saving it to later extract passwords
and anything else that moved between two computers.
    But just as packet filtering can be used to invade privacy, it can
also be used to protect privacy and security by network operators
who would otherwise be defenseless against those who are break-
 ing into their systems. One of the projects I've undertaken over
the years has been to build better packet filters to keep up with
 increasingly faster computer networks. As a result, I've been crit-
 icized by privacy rights advocates for improving technology that
 can be dangerous in the wrong hands. People have even suggest-
 ed that I was building technology for Big Brother. Obviously, like
 many technologies in the world, packet filtering can be misused
 and abused, but by itself, it's just a tool. And tools are tools are
 tools. The possibility of its misuse is not enough to dissuade me
 from developing a tool, particularly when it has such a vital role.
96                          BREAK-IN
   The first opportunity I had had to put this technology in play
against a live opponent on the Net came early in 1991 when I
received a call from Castor Fu, a former Caltech classmate.
Castor had worked with me at Los Alamos and had gone on to
become a graduate student in physics at Stanford. In January of
that year he noticed that Embezzle, one of the workstations in the
Stanford physics department, was exhibiting strange behavior.
   Investigating, he discovered that a long-dormant account
named Adrian had been taken over by an interloper who was
using it as a staging area for attacks on all sorts of government
computers. Often coming in to the Stanford network by tele-
phone, the cracker would then use the Internet to launch his for-
ays from the university's computers. Irritated, Castor went to
notify the university's computer security administrators. He
learned they knew about the attack but had decided not to take
action because they felt it was better to let the intruder continue,
and thus have some idea of what he was up to, rather than to be
left totally in the dark.
   The university's lack of concern upset Castor even more, and
he asked me to help him in his own vigilante action. We set up
monitoring software on his network, and I wrote software per-
mitting us to reconstruct the packets we captured during his raids
in a videolike presentation. Replaying the data, we could see
exactly what our Adrian interloper saw and watch all of his key-
strokes just as he typed them.
   At the time Stanford had a wide open bank of dial-in modems,
which enabled anyone to connect to the university's computers.
We ultimately discovered the intruder was a young Dutchman
who seemed to have a remarkable amount of free time to sit and
attack a variety of mostly military and government computers
around the Internet. Castor guessed he was Dutch because he
used the word probeeren, Dutch for "try," as a new password he
created for the stolen account. We also noticed that the attacks
came at times corresponding with programmers' hours in
Europe. For several months we monitored his activities and tried
to see that he didn't do anything harmful. When he did break
into other accounts on the Net, we would warn people of the
   As it turned out we weren't the only ones keeping an eye on
                        MY CHRISTMAS VACATION
      Adrian. About the same time we began tracking him, Bill
      Cheswick, a Bell Labs computer security researcher, noticed that

      someone was using Embezzle at Stanford to snoop around Bell
      Labs' system in Murray Hill, New Jersey. Rather than simply
      locking the attacker out, Cheswick decided to play with him, cat
      and mouse style. He built a phony computer, which he and his
      compatriots at Bell Labs referred to as their "jail." He set up the
      special gateway computer outside of the Bell Labs firewall
      machine and created a software "playroom" in which the intrud-
      er's every move and keystroke could be watched.
         The Dutchman whom we called Adrian was known as Berferd
      to Cheswick's team, named after the account he had comman-
      deered at Bell Labs. (The account name itself was a bit of Bell
      whimsy: in an episode of the old Dick lftzn Dyke Show, Van
      Dyke's real-life brother, Jerry, had called Dick "Berferd," because
      he "looked like a Berferd." By the same logic, the Bell Labs'
      researchers decided it was a good name for their cracker.) For sev-
      eral months Cheswick studied Berferd's activities, fed him phony
      information, and attempted to help computer-security people
      elsewhere who were trying to trace him. Meanwhile, Cheswick
      indulged in a bit of his own mischief: in the software he wrote to
      masquerade as a Bell Labs system he inserted several "wait" states
      designed to simulate a busy computer system. The Dutch attack-
      er must have frequently been left tapping his fingers on his desk
      while he waited, but he apparently never caught on.
         Occasionally the intruder did something that was outright
      destructive. Once Cheswick saw Berferd type the command
      "rm-rf / &"-perhaps the most devastating command in the
      Unix vocabulary. When issued from a root account it causes a
      computer to systematically walk through all its directories, eras-
      ing every file. Apparently Berferd wanted to cover his tracks and
      didn't care how much damage he did. Within the confines of the
      Bell Labs "jail," this command could do little harm. But Berferd's
      willingness to use it confirmed for Cheswick that he was far from
      harmless. In a paper about the attack written some months later
      Cheswick wrote, "Some crackers defend their work, stating that
      they don't do any real damage. Our cracker tried this with us
    . [unsuccessfully to erase our files], and succeeded with this com-
      mand on other systems."
98                          BREAK-IN
   Adrian and the several compatriots with whom he seemed to be
working appeared to be members of a shadowy computer under-
ground who shared information about various bugs and vulnera-
bilities in the computer systems they were attacking. Ironically,
their forays did have the benefit of revealing the poor state of
many computers that should have had real locks on their doors.
On one NASA computer Adrian tried to log in as "news"-an
account on many Unix computers to handle Usenet transactions
between different computers on the network. The computer
responded that news didn't have a password and asked him to cre-
ate one of his own!
   Another time we watched as he successfully used Robert
Tappan Morris's notorious "sendmail" bug. Sendmail is the
Internet's standard mail-handling program, and in 1988 Morris
had written a worm program that exploited a flaw in sendmail
that affected more than six thousand computers on the Internet.
That flaw had been widely known for three years, and Sun had
distributed software patching it. It was apparent that some system
administrators were just too lazy to secure their machines and
suffered the consequences.
   Yet another time, we watched as Adrian broke into the com-
puters of the Pentagon's Pacific Fleet Command and read mail.
He used a search command to hunt for all occurrences of the
word "Golf." We guessed he was actually looking for the word
"Gulf," because it was at this very point that the U.S. military
was mobilizing its forces in the Persian Gulf region. In fact late
one night Castor was stalking Adrian as he poked and pried his
way through the Internet when someone popped his head into
his office at Stanford and said, "You know, there's a war on."
   Castor looked at him blankly for a second and then said, "I
know, it's just like a war."
   The guy seemed just as baffled. Finally he said, "No, it really is
a war. The Allies just started bombing Baghdad."
   Despite the fact that Adrian was now even reading the mili-
tary's unclassified e-mail with impunity, it was hard to get the
bureaucrats in various government agencies to do anything about
 the problem. The more Castor and I watched, the more we real-
 ized that Adrian/Berferd wasn't really a skilled Unix hacker, but
simply persistent. He once sat and typed "mail -a," "mail -b,"
                   MY CHRISTMAS VACATION                         99
"mail-c," all the way to "mail -z," and then repeated the process
in upper-case, hunting for a particular vulnerability that he never
found. A lot of what he was doing also fell into the "monkey see,
monkey do" category. Since he didn't appear to know that much
and was simply copying techniques he saw, I decided to stage my
own experiment. We spent some time "teaching" Adrian new vul-
nerabilities by purposely lowering some part of a computer
defense "he was probing to briefly allow him entry. Then we
would patch it, in effect locking him out. Unaware of the ruse,
he would repeat the same trick all over the network; though he
failed everywhere else, he gave us a distinct signature with which
to identify him when he was in action.
   One night I left a telnet session running on Embezzle at
Stanford. I had telnetted to a computer at the Los Alamos
National Laboratory and then logged off the session. This left the
path to the lab still visible for Adrian to retrace, although it was
a pointer to a place I was pretty sure he would be unable to break
into. Tipped off by Castor and me, Los Alamos had been inter-
ested in Adrian, but needed an official reason to take action.
   The next day when Adrian ran the ps command to see what
programs were running on the Stanford computer he found my
abandoned telnet to Los Alamos-lanl.gov-and took the bait.
He began attempting to break into the weapons laboratory com-
puters. I called the security officers at the Lab and told them
Adrian was attacking their network. Although he was unsuccess-
ful, Adrian had become an official Department of Energy securi-
ty concern.
   Ultimately telephone tracing proved impossible, because at the
time there was no computer crime law in the Netherlands, and
the Dutch telephone company would not assist with the trace
requests from U.S. officials. However, in April, Wietse Venema,
a Dutch computer security expert, contacted security experts in
America and told them he'd tracked down a small group of
Dutch programmers who were breaking into computer systems
in the United States. He was able to identify Berferd, including
his name, address, phone number, and even his bank account
number. Around the same time I received a call from John
Markoff, the New YOrk Times reporter. We had never met, but
Markoff had heard I was monitoring the Dutch intruder. I
I 00                         BREAK·IN
described our surveillance, and on April 21 Markoff's story
appeared on the front page of the Times.

          Dutch Computer Rogues Infiltrate American
                  Systems With Impunity

                 By JOHN MARKOFF, Special to The
                        New York Times

                 Beyond the reach of American law, a group of
       Dutch computer intruders have been openly defying
       United States military, space and intelligence authorities
       for almost six months. Recently they broke into a United
       States military computer while being filmed by a Dutch
       television station.
                The intruders, working over local telephone
       lines that enable them to tap American computer net-
       works at almost no cost, have not done serious damage,
       Federal investigators say. And they have not penetrated
       the most secure Government computer systems. But they
       have entered a wide range of computers, including those
       at the Kennedy Space Center, the Pentagon's Pacific Fleet
       Command, the Lawrence Livermore National Labora-.
       tory and Stanford University using an international com-
       puter network known as the Internet.
                While the information on these systems is not
       classified, the computers store a great variety of material,
       including routine memorandums, unpublished reports
       and data from experiments. Federal officials said the
       group had tampered with some information stored on
       systems they have illegally entered.
                 United States Government officials said that
       they had been tracking the interlopers, but that no arrests
       had been made because there are no legal restrictions in
       the Netherlands barring unauthorized computer access.
       Telephone calls to Dutch Government officials in the
       Netherlands and the United States seeking comment
       were not returned.

  Although Markoff agreed not to name me, deep in the story he
included a reference to my involvement:
                   MY CHRISTMAS VACATION                             I 0I
                The Dutch group was detected last year after an
       unusually skilled United States Government computer
       researcher at a national laboratory tracked its every move
       using advanced computer security techniques. He noti-
       fied Federal authorities of the break-ins.
                The researcher has been able to make computer
       records of the intruders' keystrokes as they have electron-
       ically prowled through United States military, NASA,
       university and dozens of other computers. It has then
       been possible to play this information back and gain an
       exact picture of the computer screen as it appeared to the
       intruders in the Netherlands.

  The newspaper article and the ensuing clamor it created gener-
ated interest in my work within the government, and I eventual-
ly gave a number of presentations about Adrian and his attacks to
various agencies. As part of these lectures, I prepared a videotape
of some of Adrian's sessions so that people who weren't familiar
with computers could understand exactly how system crackers
worked, and to experience what it was like to watch a computer
screen from over a cracker's shoulder. They would be able to see
and hear, down to the bells that sounded on his terminal, what
he saw and heard in real time. I had originally planned to use the
extra soundtracks on the tape, one for snide comments on
Adrian's techniques and the other for a laugh track.
Unfortunately I never found the time or the budget.
   The Adrian incident also provided me with a useful civics les-
son. In the fall of 1991 I was in Washington, D.C., scheduled to
show my Adrian tape to researchers at the General Accounting
Office, whom Congress had ordered to investigate the break-ins.
Just before I was to begin my presentation, however, Justice
Department attorneys learned about the talk. They phoned the
GAO and demanded I not present my tape, claiming it was part
of their evidence in the case they were pursuing with the Dutch
government. While I sat and waited inside a windowless confer-
ence room, three Justice Department attorneys sped across town
in a taxi and confronted GAO lawyers, apparently concerned that
I was about to embarrass the bureaucracy. I found the whole
thing ridiculous-bureaucrats trying to coverup their lapses. In
the end I was not allowed to give the presentation until several
I 02                       BREAK-IN
months later-and then only with officials from the Justice
Department and FBI standing by.
  The uproar over the Adrian incident helped push me toward
computer security research, which in turn led me on a quest for
better tools. One of the tools I modified for my work was a
sophisticated piece of software called the Berkeley Packet Filter.
Originally written by Van Jacobson and Steven McCanne at the
federally funded Lawrence Berkeley Laboratories in 1990, it was
designed for the simple task of monitoring computer network
performance and debugging. The drawback was that it had been
created for the existing generation of computer networks. Most
businesses and research centers still use Ethernet. However,
Ethernet is an aging standard, and by 1994 I found it necessary
to create software that would be able to deal with much more
advanced computer networks, such as those that use fiber optic
cables and can reach speeds at least an order of magnitude faster
than Ethernet. Today, most large commercial online services have
internal fiber optic networks to handle the billions of bytes of
data that circulate daily between their machines. The modified
version of the BPF that I wrote was able to filter more than one
hundred thousand packets a second, even when it was running
on a Sun workstation several years old. Unlike the original BPF,
my version was designed to bury itself inside the operating sys-
tem of a computer and watch for certain information as it flowed
through the computer from the Internet. When a packet from a
certain address, or for that matter any other desired piece of
information designated by the user flashed by, BPF would grab it
and place it in a file where it could be kept for later viewing.
  I had developed my initial version of the faster BPF in the
expectation that I would receive additional research funding for
the work from the National Security Agency. The Agency had
begun supporting my work under a Los Alamos National Labs
research grant in 1991, and had promised to extend their support
for my work, but the funding was never forthcoming. I devel-
oped the tool, but after I completed the work, in early 1994, the
bureaucrats in the agency reneged on funding.
  The idea of working with the NSA is controversial in the com-
munity of security professionals and civil libertarians, many of
whom regard the NSA as a high-tech castle of darkness.
   Libertarian by inclination or by the influence of their col-
I 04                       BREAK-IN
NSA back toward the individual, and toward protecting our civil

In San Diego, while we prepared to go back on-line, our intrud-
er continued to bait us, placing a second calIon the afternoon of
December 30. When I returned to my office and played back my
voice mail, my antagonist was there again. The voice mail system
told me that the message had been left only minutes earlier, at
2:35 P.M. It began with a howling sound, a little like a yowling
cat-or was it a rooster crowing?-which then trailed off into an
odd-sounding whine.
   "Your security technique will be defeated," the message began,
in a voice that sounded as if it might be a different person from
the first caller. "Your technique is no good." A garbled sentence
followed. I listened to it over and over again, but could make no
sense of it.
   It seemed the intruder had now figured out that we were off the
Net, and he was trying to taunt us into making ourselves accessi-
ble. "These guys are being pretty cocky," I said to Andrew when
I played the message for him. "Why don't they get a life?"
Whoever it was-he, they-was trying to bait us, but I wasn't
sure why. It seemed childish. At the same time I was relieved, for
they obviously thought that they had gotten away clean, and it
seemed possible that their overconfidence might increase their
vulnerability in the future.                               ,
   On New Year's Eve the two of us were at the Center working
on the security router. We took a brief break to drive back to
Andrew's apartment where his wife, Sarah, and a small group of
friends were celebrating. The television was on, the champagne
was passed, and finally the clock struck midnight. We stayed for
a short while longer and then headed back to work. I coded labo-
riously for a couple of more hours and then at 3 A.M. went home
to sleep, my part of the routing filter essentially complete.
   I had been growing concerned about Julia, who was sounding
more depressed each time we spoke. She hadn't left Toad Hall for
days, and although she had been saying she was coming to visit
me, she had missed a couple of flights. So the next day, because
                   MY CHRISTMAS VACATION                       I 05
Julia was still arguing with John over the idea of her coming to
San Diego, I decided to fly north.
  At around 8 P.M. Andrew drove me to the airport and I left him
with a list of items to finish and loose ends to attend to. He told
me he had hopes of going back to work that evening and push-
ing on. In the end, however, our pace caught up with him, for we
had both been running on as few as four hours of sleep for five
days. Andrew went home and directly to sleep and then slept the
entire next day.
   I flew to San Jose where I picked up a rental car and drove to
San Francisco, stopping by Mark Lotter's on the way to pick up
the skiing stuff I had left there on the morning of the twenty-
seventh. My thought was to get Julia outdoors, whether it was for
a day hike or a ski in the mountains, hoping that away from Toad
Hall, she would have a chance to think about things from a fresh
perspective. She had been a close friend for a long time, and I'd
once made her a promise that if she was ever in a rut or feeling
down, I would come and we would spend time together away
from the city. She'd made the same promise to me.
   By the time I got to the city it was past 11 P.M .. Julia and I
met at our prearranged rendezvous point: Dan Farmer's flat on
the Panhandle in the Haight-Ashbury district. Dan and I had
been friends for a long time and we both had little respect for
the established computer security world. He is a controversial
security expert who would gain international attention in 1995
while he was working as a security specialist for Silicon
Graphics, Inc., a Mountain View, California, workstation
maker. The controversy was due to a security testing program
named SATAN (Security Administrator Tool for Analyzing
Networks) he coauthored with Venema, the Dutch security
expert. SATAN was designed to automatically check for wide-
ly known computer system vulnerabilities so that system
administrators would have a quick way of identifying and
assessing the weaknesses of their own networks. Hoping to
force computer security professionals out of their complacen-
cy, Dan was planning to make the program generally available
over the Internet. This meant that all the crackers would have
an easy way to prowl the Net looking for weak spots, and that
I 06                       BREAK-IN
all lazy computer system administrators who had not patched
their systems would be at risk.
   The imminent release of the final version of his program over
the Internet, scheduled for April 1995, would create intense
debate. Keeping computer security information closely held ver-
sus distributing it widely has always been a heated issue in com-
puter security circles. Dan had obviously hoped to raise the heat
by giving his program so demonic a name: SATAN.
   Dan, an ex-Marine, also has a personal style that tweaks the
more buttoned-down managerial types in Silicon Valley. Slightly
built, but with bright red curly hair that falls past his shoulders,
a penchant for black T-shirts and leather clothes, and studded
with a variety of metal objects piercing various parts of his body,
he does not match the stereotype of a computer geek. In early
1995 Silicon Graphics, in a fit of corporate cowardice and short-
sightedness, decided to fire Dan just before he released the final
version of SATAN. Several weeks later he was rehired by Sun,
SGI's competitor, but the whole affair made such a big splash in
Silicon Valley that SGI managed to lose doubly by getting bad
press and losing Dan.
   Julia and I talked late into the night; and she shared the depth
of the anger and pain between John and her. Things had gotten
much more stressful in the past week. It was becoming clear to
Julia that things weren't working, yet I started to wonder whether
there was something self-destructive in her unwillingness to end
her relationship with him.
   We were planning to hike in the Marin Headlands the next
day, but John called in the morning looking for Julia. After talk-
ing to John she seemed even more unsettled and tense. The two
of us walked over to a burrito shop on Haight Street, where I
thought we were going to get food for our hike, but Julia insist-
ed on taking lunch to John. We got the food, drove to Toad Hall,
and I waited in the car and ate my burrito. Before long both of
 them emerged to go on our hike.
   I had thought the point was for Julia to get away from the envi-
 ronment that she was feeling trapped by, but this was defeating
 the purpose of going hiking. It looked like the three of us were
 going to have to spend an awkward afternoon together, and I
wondered, why is she doing this? On the way over to Marin I
                   MY CHRISTMAS VACATION                       I 07
drove while Julia sat next to me and John sat in the backseat. The
two of them kept snapping at each other, and I finally interrupt-
ed and said, "Will you two please cool it."
   To an outsider the situation must have seemed very odd, but
jealousy was never an issue for me with Julia. Despite the deep-
ening frost in our friendship, John had long claimed that he him-
self wasn't jealous, but I had come to dismiss this as an effort to
be politically correct because I thought he was acting possessive.
It had been clear to me for a long time that nothing I did would
change the ultimate outcome of their relationship. I wanted Julia
to be able to decide for herself what she would do with her life.
If I honestly didn't feel threatened, it was because in my heart I
believed that it was Julia's decision to make, not mine.
   When we reached the Headlands we parked the car and walked
out the Tennessee ValleyTrail to the beach. Julia and I had hiked
there many times before, and I now stood by myself watching the
waves crash, and listening to the surf, while John and Julia
walked along the beach. It was foggy, windy, and chilly, adding to
the gray mood that seemed to pervade everything. At the end of
the day I went back to Dan's for the night, alone.
   Over the next couple of days, however, Julia and I spent a lot
of time together. One day we hiked near the Cliff House at a
place called Land's End, a wild spot on the edge of the ocean with
rocks, sea lions, and majestic cypress pines. We enjoyed each
other's company, and she began to get away from the situation
she had felt trapped by. Still, I could see she was afraid of antag-
onizing John, and I also realized that there wasn't much more that
I could do to help her. I still wanted to go skiing, and so I made
arrangements to catch a ride back to the mountains the next day
with Emily Sklar.
   That night I drove down to Menlo Park to visit Mark Lottor.
He met me at San Francisco Airport, where I dropped off the
rented car, and then we went hunting for nongreasy and relative-
ly healthy fast food. I had been intending to do a lot of cross-
country ski racing during the winter and so was trying to eat rea-
sonably well, even while I was traveling. But after 10 P.M. on the
Peninsula, that turned out to be an impossibility. We finally
found a Jack-in-the-Box in Redwood City. I looked for a fish
sandwich, which I will eat under duress, but it was not on the
I 08                        BREAK-IN
menu, so I ended up settling for french fries-not very healthy,
but by that time it didn't matter. When we got to Mark's place, I
was exhausted, but he wanted his own security router to protect
against the type of IP-spoofing attacks I had described to him,
and I had agreed to help him. The two of us worked into the early
   The next morning, Thursday January 5, I woke with a start at
about eleven and saw that my pager had several messages from
Emily, who lives in Palo Alto. She was panicking because she
couldn't find me and wanted to get on the road to Truckee where
she was scheduled to teach cross-country skiing that weekend.
   Emily showed up ten minutes later in a pickup truck loaded
with wood to heat the cabin. We threw my skis in the back and
headed for the Sierra. For the moment I was free from being con-
cerned about Julia, and free from my intruder. As friends, Emily
and I were comfortable talking about all kinds of things, and as
the daughter of two therapists, she had useful insights about rela-
tionships. Her suggestion in this case was that I get away from the
situation for a while-advice that seemed reasonable, and, under
the circumstances, easily practicable.
   It was raining on the way up through Sacramento, and by
Auburn, in the foothills, the rain had turned to snow. Chain con-
trol was in force, but the storm was starting to let up. We stopped
for supplies in Auburn at Ikeda's, the funky roadside fast food
restaurant/natural food grocery store that has hamburgers, which
I don't eat, but also has good milk shakes, french fries, and fresh
and dried fruit and nuts, which I do. When we got to Truckee, I
bought a pizza for dinner in town, which ended up being stone
cold by the time we got to the cabin fifteen minutes later. Because
it was late, we unloaded only our essential ski gear and comput-
er equipment. Inside, the cabin was freezing, and so I lit a fire,
then heated the pizza in the oven and ate it while Emily, who was
allergic to dairy products, made her own dinner.
   We skied only a little bit on Friday, but when the storm final-
ly passed on Saturday, we took advantage of the fresh snow by
spending the day on the trails. It was a great workout after a long
period of no exercise and little sleep, and I was grateful to be able
to forget the past two weeks.
   I hadn't been checking my San Diego voice mail regularly
                   MY CHRISTMAS VACATION                         I 09
because the phone in the cabin had been flaky when we moved
in, but when I finally did call in, there was a phone message from
Becky Bace, a computer scientist at the NSA. At that point the
agency had zero credibility with me because of its total failure to
come through with funding for a computer security group they
had encouraged me to establish. But Becky, my main contact in
the information security section of the agency, who seemed to
have been caught in the middle of an unresponsive organization,
was still trying to make it happen.
   For months she had also been trying to persuade me to come
to the Computer Misuse and Anomaly Detection (CMAD)
Conference, an annual conference on computer security and
intrusion detection that the agency cosponsored each year with
the Air Force Information Warfare Center. I'd been refusing, for
I didn't want to speak and I was fed up with dealing with the
agency. But she had kept sweetening the offer, and this year,
instead of being held on the University of California at Davis
campus, as the previous two had been, the conference would be
at the Sonoma Mission Inn Spa and Resort.
   Usually I quickly become bored with academic and theoretical
discussions of computer security, but now, in the wake of the
break-in, it seemed there was a chance to talk about something
that was more interesting, and better yet, to use our data to
describe exactly what had transpired. One of the areas of com-
puter crime detection that is still in a relatively primitive state is
methodology. For hundreds of years people have been investigat-
ing physical crimes, and while some of forensics is still a black art,
there are well-established methods for investigating crime scenes
and finding evidence. In the digital world, however, there is still
very little in the way of formal detection methodology.
   I called Becky back and she again offered an invitation to the
conference. "Why don't you just come and hang out in the hot
tub?" she said. "You don't even have to give a talk, just chat with
people." I told her I wasn't interested in freeloading, but said that
I now thought that I might be interested in attending, and speak-
ing, after all. She was delighted, and she closed the conversation
by saying she hadn't given up on the idea of the computer secu-
rity research team, adding they were on the verge of getting the
funding approved. "Yeah, sure you will," I responded.
I I0                        BREAK-IN
   But she had been prepared, at least, to pay my conference
expenses, and throw in an honorarium. I hadn't told her what I
planned to speak about, but the last thing I said was that I would
bring "a surprise."
   Later that day there was another message on the UCSD voice
mail system. This one consisted of a haunting melody, as if some-
one were playing the soundtrack to a suspense thriller I didn't rec-
ognize. It continued for thirty seconds and then ended abruptly.
It was the kind of music that is designed to make you cast a glance
over your shoulder to see if anyone is stalking you. Was someone?
Was my intruder still waiting out there for me to let my guard
down? I had no way of knowing, but he seemed to be reminding
me he was still in the hunt. If so, he would have to find an even
more sophisticated trick for sneaking into my system.

After I left San Francisco, Julia and I spoke often and it was easy
for me to see that she needed to get away from the setting she was
trapped in, so I invited her to join me at the CMAD conference.
A hot springs resort would give her an opportunity to unwind.
The conference began during the midst of the biggest California
floods in a century, particularly in Sonoma County. We arrived at
the Sonoma Mission Inn "on Tuesday January 10, in time for an
evening reception and wandered around sampling food from a
Mexican buffet and chatting with people I hadn't seen for some
time. In addition to computer security professionals there were
contingents from the military and the government intelligence
community. Being around people in the spy world is always an
odd experience, because you're never certain they are actually who
they say they are. In the rest of the world there usually are sanity
checks that tell you when you're off base, but in the classified, fan-
tasy world of intelligence, such cues often don't exist, and it's easy
for some of these people to stray far from reality.
   I introduced Julia to Blaine Burnham, who manages in the
information security section of the NSA. He shook Julia's hand
and intoned meaningfully, "I've heard a lot about you," as if
somebody who worked at a Big Brother agency might reasonably
be expected to have a dossier on all in attendance at a cocktail
party. She was immediately on guard and paranoid, and we
                   MY CHRISTMAS VACATION                       III
quickly moved on.
   The computer security world is actually a remarkably close and
inbred community, and many of the big names in the field were
at the conference. It's not a world I'm directly a part of but one
in which I like to show up every once in a while, drop a few
bombshells, and leave. The problem with conferences like
CMAD is that they're emblematic of the generally sad condition
of computer security. At any of these events the attendees are
more often than not inclined to put their heads in the sand and
refuse to acknowledge they're seeing an increasing sophistication
in the attacks.
   A lot of users who have old computer systems have decided
that, rather than fundamentally redesigning them to make them
more secure, they will purchase a black box to sit between their
computers and the outside world, giving them the illusion of pro-
tection. As a result, a lot of money is spent on building automat-
ed "intrusion detection" systems, which are derived from artificial
intelligence software that looks for what it suspects is "anomalous
behavior" on the part of users, and then sets off alarms. Much
effort is also directed toward trying to replace very expensive
security officers who actually look through records, with a pro-
gram that tries to do the same thing.
   We ran into Bill Cheswick, the Bell Labs researcher who had
monitored Adrian/Berferd several years earlier and who is a rec-
ognized expert in computer firewalls. I knew him through phone
 conversations and e-mail, but we had never actually met. He has
 a round face, curly hair, and is slightly portly without actually
 being heavy. I teased him about showing up at this posh resort,
 and he responded this was more fun than being stuck in his office
 in New Jersey in the middle of winter.
   I've always had a great deal of respect for Ches, as he's called,
who has a great sense of humor and enthusiasm. Both of us had
 grown up in the computer hacker world, where an adventure
 game called Zork was one of the first text-based games that
 emerged on mainframe computers in the late 1970s. Like many
 such games, Zork created a series of imaginary underground
 caves through which you hunted by typing commands at the key-
 board representing East, West, North, South, Up, and Down. It
 had no graphics, but that didn't really matter, since the best
I 12                        BREAK-IN
graphics are in your head. The currency of Zork was Zorkmids,
and it was Ches who had introduced me to the idea of thinking
of Zorkmids as the generic representation for money, rather than
dollars. There is usually too much emotion tied to dollars, he rea-
soned, but not to Zorkmids. People might be greedy with dollars,
but never with Zorkmids. Ches had noted (only partly in jest, I
think) that for hackers the point of large corporations was to pro-
vide enough Zorkmids so you could continue to play the game.
   Ches was also the author of the definitive computer security
text on firewalls, written with Steve Bellovin, who ironically, was
also the author of the influential paper that described IP spoof-
ing in 1989. Ches told me that in his own presentation earlier
that day he had mentioned spoofing, noting that it had never
been seen "in the wild."
   One of the other people I ran into that first evening was Tom
Longstaff, who is one of the best technical people at the
Computer Emergency Response Team, a government-funded
organization at Carnegie Mellon University in Pittsburgh. CERT,
as it is known, was set up in 1989 in the wake of the Robert
Tappan Morris Internet worm episode. It's mission is to gather
and disseminate timely information about security problems
around the Internet, but they tend to work with a bureaucratic
caution that belies the "emergency" in their name. I've always had
the sense is that he is a person who wants to do the right thing
but frequently has his hands tied by the organization for which
he works. I had tried to contact him in December after my break-
in, but we had missed making connections. k, I described the IP-
spoofing attack to him he was clearly intrigued, and I promised I
would give the complete technical description in my talk the fol-
lowing day.

I showed up downstairs the next morning with my RDI
PowerLite and my ice ax. The laptop had the notes for my talk,
and I was actually going to use the ice ax as a pointer, to drive
home my point that tools are tools. I wasn't going to actually refer
to the ax but instead hoped that its mere presence would make
people wonder what this tool was for, and maybe they'd get my
                    MY CHRISTMAS VACATION                         I I3
   I had wanted to connect my computer directly to an overhead
projector and the hotel's audiovisual system, but the conference
organizers hadn't been able to find any audiovisual equipment on
short notice. I was scheduled to speak after the first break, and
during the break I scrambled around and created transparencies
to talk from, although most of them were just screen shots of
directories or lists of commands.
   I titled my talk "What I Did During My Christmas Vacation,"
a tongue-in-cheek joke for those who knew me. I'm not a person
who attaches much importance to Christmas, which I usually
refer to as "winter break."
   Even though the topic ofIP spoofing was potentially a dry one,
I could feel the level of interest in the room pick up, because I
was the first one at the meeting to describe a real break-in and not
some theoretical computer security problem. I wanted to show
how the investigation was actually carried out, describing in
detail how I followed the trail. I noted that the attack appeared
to be scripted or automated, based on the timing of events. That
was a significant factor, because if the attack was packaged as a
program, it was likely that it could be used by people with no spe-
cial technical skills who were simply part of the network of
underground bulletin boards and Internet conferencing systems
that traffic in this kind of information. It's not that there is a
tightly knit conspiracy out there; it's just that crackers talk to each
other and they are not bound by the rules of a bureaucracy with
a cover-your-ass mentality. Their existence ensures that any new
flaw or security loophole that is discovered is known by the com-
puter underground much more quickly than by the computer
security establishment, where people don't communicate as effec-
   The fact that it had taken a great deal of detailed analysis for
me to reconstruct the break-in was something that clearly hit
home with the audience. The situation I was describing was the
type of attack that might have been going on undetected beneath
their own noses all along. The implication of my talk was that
people might have doors with huge locks, but there's a small
space between the door and the floor that bad guys can just slip
right through.
   I then played the two voice mail messages I had stored as digi-
I 14                        BREAK-IN
tal files on my notebook computer. When my computer speaker
emitted that cheesy accent, it was so distorted that it was difficult
to make out the exact words, but people got the idea that some-
one had singled me out. Many computer security types have got-
ten their heads buried so far in the sand that they've forgotten
there are real enemies out there. Me and myfriends, we'll kill you.
In front of an audience of computer professionals the voice was
chilling. There was silence in the room while I waited for ques-
   My point had been that this was a general vulnerability affect-
ing much of the Internet, because so much of it relies on address-
based authentication-electronic mail, for example. If! send you
an electronic mail message, for example, how do you know it's
really me? It's the same as if you received a postcard in the mail-
you might recognize the handwriting, but that is the only clue
you have as to whether the message has been forged. The address-
ing scheme that underlies the Internet had never been intended
to be used for authentication. It is possible to fool it in many
ways by masquerading as a trusted computer. The Internet's job
is simply to make sure that packets get from here to there, not to
provide authentication, and this attack showed the system was all
too vulnerable to subversion. You merely trust that the address is
accurate and the sender is who he or she claims to be. The attack
on my machines demonstrated that low-level Internet protocols,
the basic packaging for network communications, are wide open
and can be exploited. To guard completely against this kind of
subversion would require an exhaustive reengineering of these
 basic protocols of the Internet.
   The last question from the audience was, "Do you have any
idea who did this to you?"
   "Not really," I replied.
   After the panel a woman came up and introduced herself to me
 as Martha Stansell-Gamm. I remembered her from a law enforce-
 ment event a few years earlier. With her blond hair tied back in a
 bun and very conservatively dressed, Marty looked businesslike.
 She worked in the Justice Department computer crime unit and
 asked me if I had told the FBI about the break-in. I told her I
 hadn't and explained that in the past I hadn't had very good luck
 dealing with the Bureau. Things had gotten dropped, and so I
                   MY CHRISTMAS VACATION                        I I5
hadn't even considered calling them this time.
   "Tsutornu, I'm surprised to hear this," she said. "I'll make sure
that we're more responsive in the future." She promised to have
someone call me about the incident. Although she said she had-
n't been able to follow all of the technical details of my descrip-
tion, I got the sense that she had a sharp mind and wasn't going
to act like a bureaucrat.
   After chatting with Marty I bumped in to Jim Settle, a stocky,
square-shouldered FBI agent who had once run the Bureau's
computer crime squad. He'd left the Bureau and was now work-
ing for l-Net, a computer security contractor in the Washington,
D.C., area. In 1991, while Settle was still with the Bureau, I'd lec-
tured at one of his agent-training sessions and showed my Adrian
tapes to give them some idea of what an attack looks like. At the
time I got the impression he was slightly put off or didn't know
what to make of me. Now he was friendly and said he thought he
might have some idea who was behind the attack. I mentioned
the Oki phone software and said we had some suspicion that it
might be Kevin Mitnick, because we were pretty sure he'd tried
to steal it from Mark Lottor. But Settle doubted it was Mitnick,
because of the technical skill involved, and instead he suggested
it might be some crackers that he had heard of operating out of
   After the panel, as I spoke with members of the audience,· I
began to realize the attendees at the conference seemed to take
this vulnerability seriously, but I sensed there would be no easy
way to get their organizations to do the same thing. I had a long
chat with Tom Longstaff who acknowledged that IP spoofing was
a significant problem, but said that he thought it unlikely that he
could persuade CERT to issue an advisory because of the politics
surrounding the situation on publicizing vulnerabilities. As a
government-funded agency, CERT has always been tremendous-
ly conservative and afraid to step on anybody's toes. If they issued
an advisory warning about the danger of IP address spoofing they
would have to list the names of the manufacturers whose equip-
ment was vulnerable, a step that was politically very sensitive.
   Later, after dinner and lots of wine, I was talking to Bill
Cheswick and Marcus Ranum, another skilled computer securi-
ty specialist, about CERT's passivity, and Marcus came up with
II6                        BREAK-IN
the idea of not waiting for CERT, but going ahead and publish-
ing the details of the vulnerability on our own. I saw Longstaff
and we went over to him.
   "What would you do if we released a forged CERT advisory,
detailing the problem and warning about it?" I asked him.
   "I guess we'd have to print a retraction," he replied, and then
added with a laugh, "If it's a really good posting maybe we'd just
distribute it."
   He explained it would be very difficult to create an accurate
forgery because CERT digitally signed each advisory with an
authentication number generated using Pretty Good Privacy
(PGP), the free encryption system written by Philip
Zimmermann. I pointed out that even if everyone knew that it
was a forgery, a faked advisory would accomplish the same pur-
pose of alerting people to the problem. By the end of our con-
versation I had the feeling that Longstaff wouldn't have minded
us trying to pull off a forgery, but there was no way he was going
to encourage us.
   Julia and I spent much of the last day of the conference play-
ing hookey, because a conference at the Sonoma Mission Inn was
too good a junket to waste by spending the entire time in a dark-
ened meeting room. I did stop in for one talk being given by
Marty about how the Digital Telephony Act, which President
Clinton had signed into law the previous October, had given on-
line service and Internet providers the ability to monitor the key-
strokes of the people communicating over their systems. It was
anathema to privacy rights groups, but an important tool that
was absolutely vital for tracking intruders.
   In the afternoon we decided to go out to soak in the hot tub,
and we bumped into Marty again. At this point there was noth-
ing businesslike about her, in her one-piece blue bathing suit.
Initially she was a bit sheepish because she, like we, knew the
conference wasn't over yet, but like us, she was probably bored
stiff. She explained that she was taking a hot tub to be fully
relaxed, because she was flying directly home to Washington,
where she would have to deal with her children, including a sick
baby, and her husband, who was tired of taking care of the fam-
ily in her absence.
   It was raining lightly, which gave the whole hot tub area a won-
                   MY CHRISTMAS VACATION                        II7
derful misty quality. We chatted about the day's presentations.
Marty said that in February the Justice Department was planning
a seminar in San Diego on the legal issues relating to computer
crime. The department wanted to round up all of the assistant
United States attorneys who had been given responsibility for the
area. She invited me to the training session, and I said I would be
glad to come and speak about technology.
   Although rain began falling quite heavily, we were too content
to move and through the Sonoma mist we began talking about
my break-in. Marty couldn't understand why we weren't being
more aggressive in tracking down what we had from packet traf-
fic we'd logged from Ariel. "We had interesting pointers leading
back to places like Colorado SuperNet and Loyola University in
Chicago," I told her. "But we don't have any resources, so I
haven't been able to follow up and pursue anything." I explained
that I'd been trying to put a computer security research team
together for months but had been stalled entirely by NSA's delay
in funding us. ''I'm tired of banging my head against the bureau-
cracy and getting nowhere," I said. "I'm burned out."
   "But Tsutornu, this is a new area of law," she replied. "It's
important that we find test cases and play them all the way out."
   The conference was almost over and we all decided we should
make a final appearance as things broke up. As we were getting
out of the tub I turned and said, ''I'd love to push this thing fur-
ther but I don't have any really strong leads. Based on the data
taken I suspect that Kevin Mitnick might be behind this. But I
don't have any hard evidence. And from what I know, this kind
of an attack is really beyond his technical capability."
  Late Thursday afternoon Julia and I left the CMAD conference,
  and stopped to visit a friend of hers in Fairfax on our way back
  to San Francisco. The rain kept falling, and I kept regretting the
  snow in the Sierras that I wasn't getting to ski, but for the
  moment I was distracted by the fact that the world around us had
  been transformed into a fantastic, mist-shrouded landscape. For
  anyone who has grown up in the East, much of California always
  has this arid desert quality except in places like Marin County
  where, at the height of the rainy season, the world turns a shim-
  mering emerald green.
     On the hour-long drive down to Fairfax I was still thinking
  about the discussions I'd had with Tom Longstaff about whether
. CERT should issue a security advisory on IP spoofing. By keep-
  ing this kind of information secret from computer systems man-
  agers, whom did these guys think they were helping? It was like-
  ly that within a month the details of the break-in would be wide-
  spread in the computer underground anyway, making copycat
  attacks inevitable. The only people who wouldn't know about it
  would be those responsible for actually protecting the security of
  computers on the Internet.
     This is stupid, I thought to myself even though there didn't
  seem to be anything I could do about the situation. And, in a
  sense, it wasn't my problem. I'd gotten up in front of some of the
  nation's leading computer security experts and given a report on
                     THE PRESS DESCENDS                         II9
the exact mechanism of the attack. Now CERT was being CERT,
slow-moving and cautious, and if I was frustrated, it was in the
face of an unfortunately typical situation.
   After a Mexican dinner in Fairfax, we drove back to San
Francisco to pick up Julia's car, a 1987 Mazda hatchback, which
had sat in front of Toad Hall for several months while she had
trekked in Nepal. She followed me down to the airport to drop
off our rented Oldsmobile, and on our way back the Mazda's
engine began making an ominous grinding noise. We limped
home and spent the night at Dan Farmer's.
   In the morning I checked the oil, which didn't even register
on the dipstick, and so we eased the car to a quick-change oil
place on Divisadero. We learned later that John had loaned the
car to a number of people while Julia was away, and no one had
bothered to maintain it. She was livid. Although it had more
than 135,000 miles on it, she'd always taken good care of the
car, but now the engine sounded as if it might be permanently
   Afterward, the engine sounded sick, but was still functioning,
so we loaded the car with ski gear and took off, hoping to beat
the Friday ski traffic to the mountains. At the cabin the driveway
was full of new snow, and although it took a half hour of shovel-
ing before we could get the car parked, I was glad I was finally
going to get some skiing in.
   The next three days were a pure escape, as Julia and I skied and
talked, and spent the evenings making large dinners with Emily.
   On Tuesday night, however, I had a phone message from Tom
Longstaff. The next morning I returned his call, and he told me
 that there had been more break-ins around the Internet, using
 the IP-spoofing trick.
   "What were the targets?" I inquired.
   "Tsutomu, I'm sorry, you know that's information that CERT
 keeps confidential. I can't tell you."
   "Well, let's approach this differently," I said. "Why don't I tell
 you where I think the attacks came from."
   I checked my notes and said: ''I'll bet that your new attacks
 have come from apollo.it.luc.edu at the Loyola University in
   I was right-it matched-and I asked how it had happened.
120                         BREAK-IN
   As with the open portal between Osiris and Ariel, sorneone's
session had been hijacked. But in this case the user had actually
been sitting at his machine and could see that someone had taken
over his session. His network manager, who had been alerted that
the attack was unfolding, was able to capture the intruder's every
data packet. This information had allowed him to precisely
reconstruct the attack.
   I asked Tom ifthis new incident had changed CERT's mind, and
if they were now actually ready to put out an advisory. He would
say only that they were still considering it, but the implication was
that there might finally be enough ammunition with which to act.
   After we hung up I pondered what the new attacks might
mean. It seemed clear that IP spoofing presented a significant
vulnerability to the entire Internet community. There were thou-
sands of computer sites where trust between two computers in a
local network was a convenient, established fact, and now all of
them were at risk. I couldn't stand by any longer and wonder
when CERT was going to take action. This was something that
people should know about. But how to do it? It occurred to me
that John Markoff at the Times had asked me to call him when I
heard about things that might be significant stories. Since we had
first talked in 1991 about Adrian, Markoff and I had exchanged
information regularly about computer security and the Internet.
We had actually met face-to-face several years earlier at one of the
Hackers' conferences and discovered we both liked being in the
wilderness, and we had gone on a couple of long backcountry ski
trips together. Markoff had come to rely on me for technical
expertise in his writing about the Internet and cryptography, and
I found him a useful source of news and gossip. Markoff had told
me he wasn't interested in garden-variety computer crime stories,
only in cases in which some larger issue was involved, as in the
case of the Dutch computer cracker who was operating beyond
the reach of United States law. IP spoofing, which was an insidi-
ous weakness in the basic fabric of the Internet, seemed to be a
perfect candidate. Even if CERT was bound by its organization-
al guidelines, I figured I was free to talk about my presentation.
There were more than fifty people who had listened to my talk at
CMAD, and anyone of them could have told a newspaper
reporter about it. I might as well call one myself.
                     THE PRESS DESCENDS                       I 2I
   I reached Markoff at his office and told him about my CMAD
presentation and gave him a brief explanation of IP spoofing. I
warned him that CERT was still considering producing an advi-
sory, and that he should probably call them to ask when it was
going to be released. We agreed that he wouldn't tell the CERT
people who had told him about my paper, but since he knew at
least a half-dozen people who had been at the conference it wasn't
likely to be an issue.
   Two days later, on Thursday January 19, Marty Stansell-Gam
called to say that she had instructed the FBI to contact me. The
next day, Richard Ress, an FBI agent from Washington, left a
long message asking about the break-in and apologizing for my
past problems with the Bureau. When I returned the call, he
resumed where his message had left off, delivering a five-minute
monologue about how the FBI had run into difficulties in deal-
ing with computer crime, and how they were going to change the
way they dealt with it. He acknowledged that the Bureau realized
that it had been unresponsive to me before, and now they want-
ed to do everything possible to be more helpful in the future. It
sounded great, but I'd heard such promises before.
   Later that day Andrew spoke with Marty and then to Levord
Burns, the FBI's top computer crime field agent. I'd worked with
Burns in the past on previous break-ins, and thought of him as a
good law enforcement agent in the wrong job, for he knew little
about computers or technology. In his call Andrew had described
the particulars of our situation to Burns and then sent him a fax
outlining generally what we had learned about it.
   Markoff had been working on his story during the week and
had been negotiating with CERT over when it would be printed.
In the end, he agreed to postpone his story, after CERT officials
argued that publishing on Friday would give intruders an entire
weekend in which to experiment while network managers were
away from their systems.
   I went skiing late on Saturday. It was growing dark, and the
snow was icy, which meant very fast skiing. Only the stragglers
were still on Tahoe Donner's winding, dipping trails, and I was
alone for much of the time. As I skied it occurred to me that the
NSA would likely be dismayed by the Times article, because any
publicity tends to offend them. I stopped briefly at a warming
122                           BREAK-IN
hut and I called Becky Bace to alert her, figuring that with her
independent streak, she might be amused by the whole thing. In
a field that is notable for its lack of women, she sometimes refers
to herself as the "mother of computer security." At thirty-nine,
she seems to have been around long enough to have become a fix-
ture of sorts and to know everyone.
   I reached her at home, and as I had suspected she was actually
amused by the prospect of CMAD's getting some attention.
Here, she remarked, was this obscure academic conference that
never gets any consideration outside of a tiny community of aca-
demics, military, and intelligence people, and now one of its pre-
sentations was going to be written up in the New YOrk Times. In
the fading afternoon light I skied home wondering how things
would unfold.
   Markoff's story, which the Times first released on its newswires
on Sunday evening, was given prominent placement on the front
page of Monday's paper.

           Data Network Is Found Open To New Threat

               By JOHN MARKOFF, Special to The New
                          York Times

                  SAN FRANCISCO, Jan. 22-A Federal com-
        puter security agency has discovered that unknown
        intruders have developed a new way to break into com-
        puter systems, and the agency plans on Monday to advise
        users how to guard against the problem.
                  The new form of attack leaves many of the 20
        million government, business, university and home com-
        puters on the global Internet vulnerable to eavesdropping
        and theft. Officials say that unless computer users take the
        complicated measures they will prescribe, intruders could
        copy or destroy documents or even operate undetected by
        posing as an authorized user of the system.
                  For computer users, the problem is akin to
        homeowners discovering that burglars have master keys
        to all the front doors in the neighborhood.
                  The first known attack using the new technique
        took place on Dec. 25 against the computer of a well-
        known computer security expert at the San Diego
                      THE PRESS DESCEt\DS                             I 23
       Supercomputer Center. An unknown individual or
       group took over his computer for more than a day and
       electronically stole a large number of security programs
       he had developed.
                 Since then several attacks have been reported,
       and there is no way of knowing how many others may
       have occurred. Officials of the Government-financed
       Computer Emergency Response Team say that the new
       assaults are a warning that better security precautions will
       have to be taken before commerce comes to the Internet,
       a worldwide web of interconnected computers that
       exchange electronic messages, documents and computer

   The article went on to identify me by name and refer to my
CMAD talk. While several years earlier this kind of story would
have been buried in the back pages, if it appeared at all, the
Internet had now become big news.
   Although CERT had told Markoff that its own release would
be issued early Monday morning, the advisory document, which
told of the attacks and summarized the defensive steps that com-
puter systems managers should take, wasn't actually circulated
until 2:30 P.M. Eastern Standard Time-almost nineteen hours
after the Times story hit the newswires. That delay led to a great
deal of consternation on the part of many systems managers, who
were upset that they were first learning about Internet security
flaws from the news media. Still, the fact that a relatively arcane
matter like IP spoofing was suddenly a page-one story had a ben-
eficial side effect: usually, CERT advisories go to system admin-
istrators who typically are too busy to deal with the problems but
in this case the highest corporate executives were hearing about
the vulnerability in a form they could understand, and so there
was pressure from the top down.
   I later found out the reason for the CERT delay. The group had
decided to circulate a draft of the advisory before making a public
release about the problem and it had been confusing to many peo-
ple who weren't able to figure out how IP spoofing differed from
another class of problems referred to as "source routing attacks."
Source routing involves a similar vulnerability to IP spoofing that
124                           BREAK-IN
enables an attacker to specify a path through the Internet to
ensure that each data packet returning from a target machine is
first routed through an attacking computer. Because it is such a
well-known weakness, many traffic routing computers on the
Internet no longer permit source routing. So CERT had taken
extra time to rework the final release.
   The CERT document, which thanked Bellovin, Cheswick, me,
and three other people for contributing to understanding the
problem, began with a three-paragraph introduction:

                          January 23, 1995

           IP Spoofing Attacks and Hijacked Terminal

                 The CERT Coordination Center has received
       reports of attacks in which intruders create packets with
       spoofed source IP addresses. These attacks exploit appli-
       cations that use authentication based on IP addresses.
       This exploitation leads to user and possibly root access
       on the targeted system. Note that this attack does not
       involve source routing. Recommended solutions are
       described in Section III below.
                 In the current attack pattern, intruders may
       dynamically modify- the kernel of a Sun 4.1'x system
       once root access is attained. In this attack, which is sep-
       arate from the IP spoofing attack, intruders use a tool to
       take control of any open terminal or login session from
       users on the system. Note that although the tool is cur-
       rently being used primarily on SunGS 4.1.xsystems, the
       system features that make this attack possible are not
       unique to SunGS.
                As we receive additional information relating to
       this advisory, we will place advisories and their associat-
       ed README files are available by anonymous FTP
       from info. cert. org. We encourage you to check the
       README files regularly for updates on advisories that
       relate to your site.

  The day the Times story appeared, SDSC was flooded with
telephone calls. When Robert Borchers, the National Science
                     THE PRESS DESCENDS                        I 25
Foundation's program manager for the five supercomputer cen-
ters-SDSC's patron saint, in other words-called to find out
what was going on, the person answering the phone didn't know
who he was, and so routed him to the public relations contact,
who didn't recognize his name, either. Borchers was passed to a
series of SDSC employees who had all been instructed to say
nothing, before he finally got in touch with Sid. Borchers was not
pleased, but luckily he has a sense of humor.
   By mid-morning, SDSC had received more than forty calls
from the news media alone, and I ended up spending much of
the day in my cabin responding to reporters and people in the
Internet community anxious to learn more about the incident. I
talked by phone to the Associated Press, Reuters, USA Today, The
Wall Street [ournal; and the Philadelphia Inquirer, and bye-mail
with CNN, a news organization I've always admired because it's
difficult to slant your coverage when it is being viewed every-
where in the world. I handled these inquiries and several others,
and asked Ann Redelfs, SDSC's public relations person, to deal
with the rest. Steve Bellovin, the researcher who had written the
original paper describing IP-spoofing attacks, sent e-mail asking
for more details. But I also heard from many clueless makers of
computer security products, who, even though they had no idea
what IP spoofing was, were quite certain that their hardware or
software would protect us. It's very easy to propose solutions
when you don't know what the problem is.
   From what I'd heard, a lot of the staff at the Center were very
uptight about the situation, believing that the worst thing in the
world was to have your security breach reported on the front page
of the New YOrk Times. Sid, too, was concerned about the nega-
tive publicity, but he generally took the matter well, telling me
the only thing he regretted was not having notified Bob Borchers
ahead of time. As far as I was concerned, it was a good thing we
had been able to figure out what had happened, as opposed to the
many sites that had probably been broken into without the sys-
tem administrator's even noticing.
   I got out for a short ski in the afternoon, and when I returned
Julia and I decided that with her assistance-technical writing is
one of her talents-I should write something that described the
attack in technical detail. The cabin was filled with the gray light
of a midwinter late afternoon. Outside a gentle snow was falling,

    126                         BREAK-IN
    diffusing the lights from the ski lift across the valley, discernible
    through the living room picture window. Neither of us had both-
    ered to get up to turn on what few lights the cabin had, and I sat
    at the dining room table huddling over the soft glow of the screen
    on my portable.
       My pager buzzed, which surprised me, because it had never
    received a page at the cabin before, and in any case I had always
    assumed it was out of range. I reached over, picked it up, and in
    the gloom peered at the display.
       The digits read 911911, the emergency code, repeated.
       "Odd," I said to Julia, showing her the tiny screen. I set it
    down. Seconds later it buzzed again, and the same six digits
    appeared again.
       Why would someone page me with 911? We looked at each
    other, through the deepening gloom. Almost no one had my
    pager number, which changes frequently, but among my stolen
    files was a back-up "image" of the memory inside my cell phone,
    which contained a directory that included my pager number. Was
    this another message from the same person who had been leaving
    me the cryptic voice mail? I set the pager on the table, and
    watched as every half-minute or so it buzzed on the hard surface
    like a rattlesnake for ten or fifteen seconds, each time producing
    that same eerie string of digits, 911911, as if someone were warn-
    ing me to call for help. Here we were, deep in the Sierras, in a
    remote cabin: If they knew my pager number, what else did they
       I called PageNet, the pager company to which I subscribed,
    and told them someone was harassing me, and asked them to
    attempt to trace the phone calls. Julia and I watched the pager
    continue to jitter on the table, until I finally reached over and
    shut it off. I didn't feel fear, just the spooky sense that someone
    could now really badger me if they wanted to.
       We didn't get started on our writing project until about 10 P.M.,
    after dinner, but by 3:30 in the morning we had a detailed doc-
    ument, which I planned to post on Usenet. Unlike the CERT
    advisory, which carefully left out any of the names of organiza-
    tions who had suffered break-ins, I was not going to change the
    names to protect the guilty.
1-'                    THE PRESS DESCENDS                       I 27
     The Usenet, which predates the Internet by a few years, began
  as an anarchic messaging system for many of the world's Unix-
  based computers, which were originally linked mostly by regular
  telephone lines and modems. From the beginning, the Usenet
  was organized into news groups where people could post and
  read messages pertaining to the group's designated subject matter.
  Now there are more than twelve thousand different groups
  where people can discuss every subject imaginable. I planned to
  post my report to three news groups where security issues were
  discussed regularly-comp.security.mise, comp.protocols.tcp-ip,
  and alt.security-when I got back to San Diego.
     The title of my message was "Technical details of the attack
  described by Markoff in NYT." It began: "Greetings from Lake
  Tahoe. There seems to be a lot of confusion about the IP address
  spoofing and connection hijacking attacks described by John
  Markoff's 1/23/95 NYT article and CERT advisory CA-95:01.
  Here are some technical details from my presentation on 1/11/95
  at CMAD 3 in Sonoma, California. Hopefully this will help clear
  up any misunderstandings as to the nature of these attacks." The
  posting then detailed the attack blow by blow, starting with ini-
  tial probes from toad.com, the computer in John Gilmore's base-
  ment, and winding up with the hijacking of Osiris.
     While Julia and I were working I called Andrew in San Diego.
  We talked about his work improving our software security
  perimeter, and his conversations with Levord Burns of the FBI,
  and I told him about the flurry of 911 pages.
     There was a long pause on the other end of the line.
     Finally, Andrew said quietly: "Tsutornu, that was me. I just
  added you to the list of alert phone numbers. There was a bug in
  the filtering router code I was setting up that made it trigger
  when it shouldn't have." He had been building alarms in our
  defensive software to automatically send pages if a break-in was
  attempted. The digits 911911 were his unmistakable way of indi-
  cating an emergency.
     In one sense it was a relief to know the reason my pager had
  been behaving like a rattlesnake, but I was exasperated, and
  thought to mysel£ Andrew, that's why I'm the adviser, andyou're the
  graduate student.
I 28                        BREAK-IN

Early Tuesday morning, after getting about three hours of sleep,
Julia and I drove to the Reno airport in miserable weather. She was
 taking the United Express 9:35 flight to San Francisco and I was
flying Reno Air to San Diego at 11:20. It had been an enjoyable
 ten days of working, playing, and sharing each other's company,
but now we had to make our farewells at her departure gate. She
was heading back to see John because she had promised to speak
 to him that morning. But she was reluctant to go and told me she
was feeling scared, yet she felt she needed to honor her commit-
ment. I had no idea when we would see each other again, and I
felt concerned, since the last time she had gone to Toad Hall she
had been miserable and it had been very hard for her to get away.
.As she walked to her plane, I waved to her through the large plate
glass window and watched her plane taxi onto the runway.
   After her flight had left I went to check in for my own flight and
while I waited I pulled out my RadioMail terminal, the wireless
modem connected to my Hewlett Packard 100 palmtop comput-
er that allows me to send and receive electronic mail in every
major city in the country. It took almost an hour for the
RadioMail unit to suck in the unusually large number of messages
that had been queued for me. One was from a mailing list that I'm
on for a bunch of my former Caltech classmates, and it mentioned
that there was an article about a Shimomura on the front page of
 USA TOday. Then somebody else sent me a note saying, "Are you
the same guy?" I sent back a note: "Yeah, I'm the same guy, and
finding your name in the New York Times like this is a lousy thing
to have happen in the middle of your ski vacation."
   When I got to San Diego I made an appearance at a meeting
of the senior fellows at SDSC, and then spent some time return-
ing more calls from reporters. Andrew meanwhile was dealing
with campus security people and had also contacted the FBI to
see if we could get a trap-and-trace order on my office telephone.
We figured that our intruder was likely to call again, and if he
did, we wanted to know where he was calling from.
   On Tuesday afternoon I realized I'd been sitting all day and
decided that if I couldn't ski, at least I'd go skating, and went for
a 25-kilometer skate around Lake Miramar, where an informal
group of in-line speed-skaters gathers every evening. Whatever
kind of privacy I'd enjoyed in the past seemed to be vanishing, but


                          THE PRESS DESCENDS                        I 29
     among this pack of two dozen skaters I could remain anonymous.
        When I got home later that night, I decided to add one final
     touch to my break-in report before posting it to Usenet. I had pre-
     viously made digitized files of the "We'll kill you" and the "Your
     technique is no good" voice mail messages left by my intruder,
     and now I was going to make them publicly available. SDSC
     operated an Internet site permitting ftp computer file transfers.
     The site contained free software available under a variety of head-
     ings and I had Andrew put the voice mail messages in a directory
     that we named pub/security/sounds/-a not so veiled reference
     to Usenet distribution categories of similar redeeming value
     like alt.sex.sounds. He named the two files tweedledee.au and
     tweedledum.au, the "au" indicating audio files for downloaders'
     listening pleasure. I added a note about the files in my "Technical
     details" report, and finally posted it to Usenet at 4 A.M.
        Tweedledee.au and tweedledum.au enjoyed a brief popularity
     on the Internet audio charts. The San Jose Mercury News created
     a World Wide Web page link to the files from their Mercury
     Center Internet service, and Newsweek included quotes from
     them in its article about the break-in. I was guessing that my
     intruder, like many people, read the newspapers and news maga-
     zines, and would probably be feeling pretty pleased with himself
     about the attention his handiwork was receiving. Maybe I could
     bait him into calling again, and ifhe did, maybe we could spring
     our telephone tracing trap on him.
        I had been playing phone tag with David Bank, a San Jose
     Mercury News reporter who covered telecommunications, and I
     finally talked to him in the afternoon. He told me he was hav-
     ing trouble finding my Usenet posting, so I sent him a copy via
     e-mail, and thought nothing more about it. Several days after he
     read my report about the initial toad. com attack, he called John
     Gilmore to ask about it. John told Bank I'd actually been at Toad
     Hall when the attack on my San Diego machines occurred. Bank
     began to investigate a scenario in which, as part of my funding
     dispute with the NSA, I had broken into my own computers to
     generate material for the CMAD talk. The premise was that a
     phony attack of my own invention would give me an opportu-
     nity to show off my skills to my potential sponsors. The flaw in
     the theory was that everyone knew that I was friends with Julia

      I 30                       BREAK-IN
      and John. Why would I be stupid enough to fake an attack, then
      publish information that would point almost directly back to me?
      But Bank, who was a persistent former metro reporter at the
      Mercury News, began to dog my trail, calling all kinds of people
      who knew me, in an effort to prove his hypothesis.
         The real trail lay elsewhere and gradually the first indications
      of it were starting to appear. On January 17, while I was at
      Truckee, SDSC received an e-mail from Liudvikas Bukys, a com-
      puter system administrator at the University of Rochester. The
      last paragraph of his note warned that the Center might have
      severe problems if it didn't already know about the break-ins, for
      security managers at Rochester had found pointers back to my
      network while they were investigating a break-in of their own.
      Andrew had talked to the Rochester group and then to security
      staff at Loyola University as well, which had had a similar break-
      in and had also been notified by Rochester.
         Andrew had called me at Truckee and told me that somehow my
      files had been transferred to a computer at the University of
      Rochester from Loyola so that the security managers there could
      examine them. The Rochester administrators were concerned that
      they had lost some source code from Silicon Graphics's IRlX oper-
      ating system during their own break-in. However, when the com-
      puter administrators at Rochester examined the stolen files that
      had been found at Loyola, they realized they were my files. Andrew
      had initially been confused, and for a while we believed that my
      files had been stashed at Rochester by the intruders as well, which
      wasn't true.
         We also learned that over the weekend somebody else had dis-
      covered that my files had made their way to Rochester, possibly
      because of the newspaper articles that reported the break-in there.
      Whoever it was had again broken into the Rochester machines
      using the same IP-spoofing attack, stolen the files again, and then
      deleted them. The spoofing attack was successful at Rochester a
      second time because although the network administrators at the
      university had configured the university firewall to defeat such
      attacks, they had unfortunately made a configuration error when
      they were changing their router's software.
          On Wednesday, Andrew and I looked at the files that had been
      retrieved from Loyola University. There was one interesting clue:
                          THE PRESS DESCENDS                           I 3I
    whoever hid the stolen data on the University computer had gone
    through my files to see what he had taken. One thing that caught
    our attention was that a digitized picture of Kevin Mitnick had
    been pulled out of my compressed files and stored separately.
    Why, I wondered aloud, of all the material had the thief left the
    photo of Mitnick lying about?
       Was Mitnick our intruder?
       "Nah," both Andrew and I said almost simultaneously-it was
    too obvious. Besides, in a March break-in into SDSC's comput-
    ers, an intruder had planted information to make it look as if the
    attack had been performed by someone else, and so we were both
    extremely wary about attributing too much significance to such
    an obvious clue.
       I'd thought the press attention would begin to die away, but
    instead it continued to build up throughout the week. That after-
    noon, two reporters working on a story for the Gannett papers in
    Rochester had listened to the recordings at the ftp site and called
    me at SDSC, asking how it felt to receive a death threat. I told
    them I didn't take it seriously. On Thursday, several publications
    sent photographers, essentially wasting the entire day. Adding to
    the media frenzy was the fact that the new university athletic cen-
    ter next door to SDSC was being dedicated that day, and Hillary
    Clinton was attending the ceremony, surrounded by reporters
    and camera crews. A Newsweek photographer showed up at my
    office with an array of photo gear and lots of lenses and special
    gels, and told me his editors had instructed him to shoot pictures
    "in the style used by Wired," the San Francisco magazine about
    Net culture that is known for its bizarre graphics.
       We took pictures at the Center for a while and then the photog-
    rapher suggested we drive up to Torrey Pines State Park and shoot
    pictures there. But we arrived just after the gate had been closed,
    and so were forced to park outside, where the photographer posed
    me on a clump of rocks. I'd been a photographer in high school,
    and now for the first time I realized what it was like to be on the
    other side of the lens, feeling entirely silly as I sat there pretending
     to play with one of my portable computers. A group of people
    walked by and stopped to watch. I heard one ask, "What"s going
     on here? Are they shooting a computer ad?"The published "photo"
    was obviously the art director's idea of being avant-garde; for it was
I 32                        BREAK-IN
a bizarre picture of me sitting cross-legged with my laptop, with an
inset picture of me erupting from my own head.
   That evening Julia called, and although things were unpleasant
at Toad Hall, she sounded better. We had only spoken briefly dur-
ing the last couple of days because I had been so busy. I told her
about my adventures in.the face of the media onslaught, and how
silly I'd felt playing model for the photographers earlier in the day.
   "This time, things started out badly and got worse," she said.
"I can see that what is going on here is bad for me, and that I
need to make some hard choices."
   This was something that I hadn't heard before from Julia.
Previously she had been fearful of leavingher familiar surroundings.
   "I would like to be with you and I know that means leaving
John," she said a quiet voice. "But this is going to be very hard
because we've been involved for a long time."
   She still wasn't certain how she was going to make the break,
but I was elated.
   ''I'll do anything I can to help you," I told her. We had missed
each other and agreed to meet as soon as possible.
   The following night I went on the Friday night skate that con-
venes each week at 7:30 P.M. at a bike shop in Mission Beach.
There are usually only fifteen or twenty people along-a much
smaller version of the Friday "Midnight Rollers" night skate in
San Francisco, which often attracts more than four hundred peo-
ple and which I try not to miss when I'm in the Bay Area. The
San Diego skate, a leisurely social event called the Dinner Roll,
usually goes for about 20 kilometers over a course that ends up in
a part of the city with lots of restaurants.
   At around 9:30 we had just started the return leg of the route
when the cell phone in my fanny pack rang. My phone has a
combination earplug-microphone so I can converse and skate at
the same time, and I listened as Sid told me that he had checked
the New York Times service on America Online, and found an
early copy of Markoff's article.
   "You're not going to like what it says," Sid told me, and began
reading it aloud, with sarcasm: "It was as if the thieves, to prove
their prowess, had burglarized the locksmith," he began. "Which
is why Tsutomu Shimomura, the keeper of the keys in this case,
is taking the break-in as a personal affront-and why he consid-
ers solving the crime a matter of honor."
                      THE PRESS DESCENDS                         I 33
   I couldn't help smiling at Markoff's melodramatic prose as Sid
continued reading and I kept rolling: "Mr. Shimomura, one of
the country's most skilled computer security experts, is the per-
son who prompted a government computer agency to issue a
chilling warning on Monday. Unknown intruders, the agency
cautioned, had used a sophisticated break-in technique to steal
files from Mr. Shimomura's own well-guarded computer in his
home near San Diego. And the stealth and style of the attack
indicated that many of the millions of computers connected to
the global Internet network could be at risk. There have been at
least four other known victims so far, including computers at
Loyola University of Chicago, the University of Rochester and
Drexel University in Philadelphia."
   The article quoted me at several points, including: "Looks like
the ankle-biters have learned to read technical manuals. . . .
Somebody should teach them some manners."
   When Sid finished, I told him I actually thought the article was
reasonably entertaining. But he was upset, and he feared that the
effect of the article would be to make SDSC a huge target.
"Tsutomu, this is deliberately inflammatory and challenging," he
fumed. "You're clearly trying to provoke them."
   As I headed back, we talked about the article for more than ten
minutes, concluding that we would have to wait and see what
   Later, after midnight, I went down and picked up a copy of the
 Times at the Circle K, a local convenience store about a five-
minute drive from my house, to see for myself how provocative
the article actually was. As I read it over, I realized Sid was right,
it was inflammatory, but it was still entertaining.
   The next morning I was up fairly early because my landlord
was thinking about selling the house and came by with his wife
to look things over. It was the first low-key morning since the
CERT story had appeared. Life seemed finally to be returning to
normal, and I was looking forward to a calm weekend.
   I was standing in my living room talking to my visitors,
Markoff called, and I told him I'd call right back. Five minutes
later I phoned him at the Times San Francisco bureau.
   "I've got bad news," he began. "Your stolen files have been
found at the Well."
On Friday night, January 27, when he logged into his account on
the Well, a popular computer conferencing service in the Bay
Area, Bruce Koball had made a puzzling discovery.
   Koball, a Berkeley-based software designer, is an organizer of
the annual Computers, Freedom and Privacy conference, and the
Well had given Koball an additional free account to help in his
preparations for the 1995 gathering. Because he hadn't used the
CFP account for several months, he was surprised to log into his
personal Well account on Friday evening and see a message from
the Well's system-support staff warning him to remove the 150
megabytes of material currently stored in the CFP account, or
they would delete it for him.
   Such messages are common at the Well, whose staff routinely
runs an accounting program to uncover "disk hogs"-people
who take up inordinate storage space on the service's comput-
ers-and Koball was simply one of several Well subscribers to
receive such a warning that day. But since Koball had barely used
the CFP account, the message made no sense until he looked at
its directory. The account had clearly been taken over by an
intruder and filled with mysterious compressed files and batches
of electronic mail:

well% Is -1
total 158127
                         KOBALL'5 FIN D                          I 35
-rw-r--r--    1 cfp    128273 Dec 26 23:02      bad.tgz
-rw-r--r--    1 cfp    547400 Dec 26 23:07      brk.tar.Z
-rw-r--r--    1 cfp      6620 Dec 26 23:07      clobber.tar.Z
-rw-r--r--    1 cfp      2972 Dec 26 23:07      clobber.tgz
-rw-r--r--    1 cfp       734 Mar 14   1991     dead. letter
-rw-r--r--    1 cfp    704251 Dec 26 23: 11     disasm.tar.Z
-rw-r--r--    1 cfp   4558390 Dec 26 23:31      fi1e.941210.0214.gz
-rw-r--r--    1 cfp   1584288 Dec 26 23:39      file.941215.0211.gz
-rw-r--r--    1 cfp   2099998 Dec 26 23:47      file.941217.0149.gz
-rw-r--r--    1 cfp   1087949 Dec 27 10:09      kdm.jpeg
-rw-r--r--    1 cfp    275100 Dec 27 10:09      kdm.ps.Z
-rw-r--r--    1 cfp   1068231 Dec 27 10:10      mbox.1.Z
-rw-r--r--    1 cfp    869439 Dec 27 10:10      mbox.2.Z
-rw-r--r--    1 cfp    495875 Dec 27 10:10      mbox.Z
-rw-r--r--    1 cfp     43734 Dec 27 10:10      modesn.txt.Z
-rw-r--r--    1 cfp   1440017 Dec 27 10:11      newoki.tar.Z
-rw-r--r--    1 cfp    999242 Dec 27 10:12      okitsu.tar.Z
-rw-rw-rw-    1 cfp    578305 Dec 28   09:25    stuff.tar.Z
-rw-rw-rw-    1 cfp 140846522 Dec 27 11:28      t.tgz
-rw-r--r--    1 cfp    146557 Dec 27 11 :28     top1evel.tar.Z
-rw-r--r--    1 cfp   3967175 Dec 27 11:31      tt.Z
-rw-r--r--    1 cfp       307 Dec 20    1990    xmodem.log
-rw-r--r--    1 cfp    187656 Dec 27 11:31      ztoo1s.tar.Z

   The directory listing showed the total amount of disk space the
files occupied as well as their names, modification dates, and
other details. Among the files were three named "mbox," stan-
dard Unix nomenclature for the file that holds a user's mail.
When Koball looked in some of the mail files, he found that
all the messages were addressed to the same party:
   Although Koball and I had attended a number of annual gath-
erings of computer industry pioneers called Hackers' Conferences
together, the name Tsutomu didn't immediately click. He was baf-
fled and still wondering what to do about his discovery when he
heard the next day's edition of the New YOrk Times land on his front
doorstep later that evening, as it does before midnight at many
homes in the Bay Area. Koball brought in the paper, browsed
through it, turned to the front of the business section, and there it
was-MarkofF's article and a photo of me sitting in the the opera-
tions room of the San Diego Supercomputer Center.
   Early the next morning Koball phoned Hua-Pei Chen, system
I 36                       BREAK-IN
adminstrator for the Well, which is based in Sausalito in Marin
County. He reported his discovery, told her about the connection
to me, and asked her to delete the files and kill the CFP account.
   A short time later Koball received a call from his friend John
Wharton, a freelance chip designer and a principal organizer of
the elite Asilomar conference on microprocessor design attended
each year in Monterey by many of the "graybeard" pioneers in the
semiconductor and personal computer industries. Wharton, who
was driving up Highway 101 to the Cow Palace in San Francisco
on his way to a model-railroad show, wanted to know if there
would be a demonstration of a digital model train sound-effects
system that Koball had developed with Neil Young, the rock star,
who is also a model-train fanatic.
   Koball is a pioneer in an industry that is injecting computer
"intelligence" into all sorts of consumer products, and has devel-
oped software for particularly neat gadgets like the Avocet
cyclometers used by serious bicyclists and altimeter watches
favored by climbers-myself included. (When Julia was in Nepal
she'd noticed that Koball's watch is the definitive status item
among the Sherpas who accompany Western climbers in the
   Koball told Wharton that no, his model-train sound system
wouldn't be on display, but then he quickly turned the conversa-
tion to his Well discovery. Wharton was fascinated, and as the
two discussed the matter, he agreed that killing the CFP account
seemed the reasonable approach. After the call ended Wharton
began to wonder if either of them really knew enough to reach
such a conclusion, and it occurred to him that a good person to
consult would be his friend Marianne Mueller, a systems software
programmer at Sun Microsystems who was far more knowledge-
able than he about Unix, the Internet, and security issues in gen-
eral. He tried phoning her home and work numbers, but was
unable to reach her.
   I had in fact encountered both Mueller and Wharton a year
earlier in Las Vegas at the annual convention of the computer
underground, called Defcon, a bizarre gathering of anklebiters,
telecom industry security people, and undoubtedly a few cops,
that I had somehow let Markoff drag me to. One of the few high
points was a talk Mueller gave about female hackers, in which she
                         KOBALL'S FIND                         I 37
introduced her own digitally outfitted version of a Barbie doll,
which she called Hacker Barbe, deliberately misspelling the name
to avoid copyright problems with Barbie's maker, Martel, Inc. It
was all quite funny, though the humor was lost on the teenage
and college-age boys attending the conference, who seemed pri-
marily interested in juvenile pranks like defeating the hotel's
microprocessor-controlled door locks.
   By this time Wharton was approaching the turnoff for the San
Francisco airport, and he remembered that Mueller had men-
tioned on the previous day that she would be seeing a friend off
to Tokyo around noon on Saturday. He took the airport exit,
drove to the top of the parking garage, spotted Mueller's MR2 in
the international departures area, and was able to park right next
to it. As he ran into the terminal Wharton saw from a departures
monitor that a JAL flight was just leaving. He stopped in the con-
course, and began to scan the crowd.
   Wharton, with his long, bushy, graying hair, wire-rim glasses,
and Birkenstocks, hadn't been standing long in the airport con-
course when he spotted Mueller, dressed in black motorcycle
leathers and a Cypherpunks T-shirt, and ran up to her. Milking
the clandestine possibilities of this encounter in an international
air terminal, Wharton swore Mueller to secrecy before telling her
what he'd just heard from Koball.
   "Nothing should be done to the files!" Mueller said, alarmed.
She insisted that they not be erased or altered in any way, for any
changes might tip off the thief that his presence had been dis-
covered. Instead, she explained, the Well's security staff should
put monitoring software in place and keep a log of everyone who
tried to access the files. Wharton, still enjoying this cloak-and-
dagger plot, but now more conscious of the seriousness of the
matter, called Koball back with his cell phone.
   "Red Dog Leader," he began, when Koball answered. "This is
Cosmic Whiner. I just reconnoitered with Cyber Mama; she says
you should leave them files alone]"
   After speaking a short time later with Mueller over a regular
phone, Koball called the Well back to countermand his prior
instructions. As it turned out, worried Well officials had already
discussed the situation among themselves, and had decided not
to take any action yet that might risk alerting the intruder. Koball
I 38                       BREAK-IN
then phoned Markoff, who in turn called to alert me and give me
Koball's number in Berkeley.
  It was nearly noon when Bruce Koball and I spoke. He
described the files he'd found the previous evening in the CFP
directory, and read me the access dates, not all of which were use-
ful because he had already read some of the files. But there was
no question that they consisted of material stolen from Ariel in
December-the programs I'd written, the extraneous free soft-
ware that had been pointless to steal, and, worst of all, megabytes
and megabytes of my mail, which felt like a tremendous invasion
of my privacy.
   He briefed me on his conversations earlier that morning, with
Pei at the Well, with Wharton and Mueller, and then with Pei
once again. It was a great plot, but I wasn't interested in any spy
stuff no matter what Markoff's article might have said about my
considering "solving the crime a matter of honor." It was my elec-
tronic mail that was being scattered all over the Internet, and I
told Koball I wanted the files erased. He gave me Pei's cell phone
number, and I called to tell her the same thing.
   Pei explained that the Well officials were uncertain about how
to proceed, since they were concerned that erasing the files would
not only give notice to the person who had taken over the CFP
account, but might also provoke the intruder into some sort of
   "I want those files off your machine," I said again. From what
Koball had told me, the intruder had set the "permission" para-
meters on the CFP account to "world readable," meaning that
anyone with access to the Well could look at the files it con-
tained. I wasn't comfortable with making my mail public under
any circumstances, and I especially didn't want any of my per-
sonal or professional files becoming the popular literature of the
computer underground. In any case, I also understood the Well's
qualms, so I began talking Pei through a set of steps that would
make it look as if Koball had simply responded to the disk hog
warning and cleaned the files from the CFP directory without
taking note of their content.
   I didn't know very much about the Well, at that point. Several
months earlier, when I had been consulting at Sun Microsystems,
I had briefly met the Well's owner, Bruce Katz (pronounced
                          KOBALL'S FIND                         I 39
 cates), who was in the market for new computers. I had only a
 hazy recollection that the Well was an electronic hangout for
 everyone from hackers' conference regulars to Dead Heads with
    Pei, for her part, seemed rather hazy about the Well's security ,
 vulnerabilities. Like many people, she had been reading about the
 San Diego break-in, and as of that morning she knew that at least
 one Well account had been cracked, but she hadn't connected the
 two events.
    ''Ah, that makes a lot more sense," she said, after I explained to
 her how my files had ended up on her system. She acknowledged
 that the Well knew it was probably under-equipped to handle
 this problem, and said that during that morning's frantic round
 of phone conversations, one of the Well's board members, John
 Perry Barlow, a Grateful Dead lyricist and one of the founders of
 the Electronic Frontier Foundation, had suggested that the Well
 invite me to Sausalito to help them.
    I was scheduled to speak at a computer conference in Palm
 Springs the following week, where I had arranged to meet Julia,
 and told Pei I was already overcommitted. But how about, I sug-
 gested, if I had my student Andrew Gross fly up to Sausalito and
 take a look around?
    I spent several hours making the arrangements. Andrew and his
.wife, Sarah, a doctoral candidate in chemistry at UCSD, were in
 the process of moving to another apartment because the mold in
 their student flat had become intolerable, even to someone with
 Sarah's fascination with organic matter. They agreed, however,
 that he could spring free by Tuesday for a flight north. I next per-
 suaded Sid to continue Andrew's Supercomputer Center salary,
 and even kick in some expense money, while he was in the Bay
 Area. I still wasn't intent on tracking the intruder, but I thought
 that if Andrew went up to the Well we might be able to discover
 additional useful information about the Ariel break-in and what
 had happened with my files afterward.
    Koball had by now e-mailed me the file-creation and access
 times for the contents of the CFP directory (and added a message
 saying he'd been glad to see that I was wearing one of his Avocet
 Vertech altimeter watches in the Times photo). Once I knew
 when the files had been created I was able to determine that
I 40                        BREAK-IN
copies of my material had been deposited at the Well within
twelve hours of their removal from Ariel. While it was possible
that the files could have made one or even several stops elsewhere
on the Internet before landing at the Well, the timing suggested
that the Well's cracker was someone with a fairly close link to the
person who had stolen my files-or was the thief himself.

On Wednesday afternoon, before I left for the Palm Springs con-
ference, I received a phone update from Andrew, who had gotten
to Sausalito the previous evening. Anticipating the arrival of a
computer commando, the Well had rented him an imposing Jeep
Cherokee, with which Andrew was quite taken, since his own car
in San Diego was a thirteen-year-old Honda Accord. Otherwise
he was frustrated: so far his entire stay had been taken up in
meetings with Well officials about nondisclosure agreements they
wanted him to sign before letting him peer into their system; he
had spent no time yet actually looking at their problems. Because
he is typically much more accommodating and diplomatic than
I, Andrew's unhappiness struck me as a bad sign of how muddled
things might be at the Well.
   After flying to Palm Springs, I checked into the Westin Mission
Hills Resort, where the conference sponsors were picking up
expenses for Julia and myself. It was late afternoon, and Julia had-
n't arrived yet from San Francisco, having missed her flight. She
had become embroiled in another agonizing discussion with
John; who vowed that if she went to Palm Springs with me, their
relationship was over. She had told him that she was going any-
way, but she didn't want to end things in anger. As a compromise
she had agreed that they would meet the following weekend to
say goodbye.
   I went off to the speakers' reception, leaving a note at the front
desk for her, and she finally caught up with me at the conference
dinner at around 9 P.M. She was stressed out and exhausted from
having spent hours on stand-by in Los Angeles until she could get
a flight to Palm Springs. But she had made it, and we were
delighted to see each other.
   This event was a Vanguard Conference, one in a series of sem-
inars that the Computer Sciences Corporation, a high-tech con-
                         KOBALL'S FIND                         I 4I
sulting firm, conducted throughout the year for high-ranking
corporate executives responsible for information technology at
their companies. The attendees included representatives from a
long list of businesses like AT&T, American Express, Federal
Express, Morgan Stanley, and Turner Broadcasting.
  The speaker list was also impressive and varied, and included
Bill Cheswick, the Bell Labs computer security expert; Whitfield
Diffie, father of a widely adopted technique for computer priva-
cy called public-key cryptography; Clifford Stoll, the itinerant
astronomer who in the mid-eighties tracked down some young
German computer vandals and then turned the story into a best-
selling nonfiction book, The Cuckoo's Egg, Mitchell Kapor, the
founder of Lotus Development and cofounder of the Electronic
Frontier Foundation; and Nicholas Negroponte, founder and
director of the MIT Media Laboratory.
  I had been invited as a last-minute replacement for another
speaker by one of the conference organizers, Larry Smarr, direc-
tor of the federally financed National Center for Supercomputer
Applications in Illinois, one of SDSC's sister organizations. I had
assumed that my sudden celebrity had a lot to do with the invi-
tation, though I wasn't sure whether to be flattered when I
learned that I was actually taking the place of Mark Abene, a con-
victed computer felon from New York, better known as Phiber
Optic, who had to cancel when his parole officer refused to let
him leave the state. But I felt better when I ran into Bill Cheswick
at the speakers' reception; Ches joked that he and I now seemed
to be regulars on the computer security rubber-chicken circuit.
   People kept coming up to me remarking, "I saw your picture
in the paper," so often that I developed a standard response: "It's
better than having your picture in the post office."
  My well publicized break-in had engendered a striking degree
of paranoia among the corporate types in attendance, many of
whom were responsible for their companies' networks. They had
apparently concluded that if "one of the country's most skilled
computer security experts" could get attacked with impunity,
then they were obviously even more vulnerable. "It's a sad day
when you have to learn about security problems first in the pages
of the New York Times," someone from Morgan Stanley, the big
investment bank, had written me in an e-mail message after
142                         BREAK-IN
Markoff's CERT story appeared early the previous week. Now
executives at the conference were approaching me and asking if I
would be willing to come to their companies to do a security
review and to consult.
   On Thursday, I gave an only slightly more polished version of
my CMAD presentation, demonstrating what an IP-spoofing
attack looked like "in the wild," but I'm afraid my attempt to
underscore the complexities of real computer crime only had the
effect of bewildering many of my listeners. That Friday, though,
on a panel with Ches, I took the audience through my Adrian
video from the 1991 government and military break-ins, and the
group not only followed along but seemed to take satisfaction in
seeing a case where the "good guys" had been able to detect, and
then contain, the villains.
  After the talk Julia and I skated on the bike paths and streets of
Palm Springs, marveling at the immaculately maintained lawns
here in the middle of the desert. By the time of the outdoor
reception Thursday evening, serenaded by a country-and-western
band, we had finally started to have fun and were not looking for-
ward to leaving the next day. And yet the nearby snow-capped
San Jacinto peaks served as reminders that we would be heading
back up to the Sierra Nevada, to pick up the skiing we had inter-
rupted nearly two weeks before.

Early Friday afternoon, Larry Smarr and I became engrossed in
discussing the possibility of a joint computer-security research
project between SDSC and Smarr's Center. Too much of what
passed for computer security, we had decided, was simply a defen-
sive duck-and-cover posture; we wanted to develop a much more
aggressive model for confronting the enemy, drawing upon mili-
tary war-game theory, and discovering the extent to which it could
be applied to the electronic realm. If computer intruders were rou-
tinely hunted down and identified it would dramatically lower the
incident rate.
   Losing track of the time, I realized that I had to find Julia, who
I discovered talking shop with a group of systems professionals. I
pulled her away so we could make our commuter flight to Los
Angeles, where we were to catch a connecting flight to Reno. As
                         KOBALL'5 FIND                          I 43
it was, we barely made it, getting delayed in Palm Springs rush-
hour traffic, and ending up charging through the terminal, a light
brigade of portable computers, skate tote bags, carry-on luggage,
and ski gear.
   The next morning we drove to Mount Rose, 20 kilometers
southwest of the Reno airport for a ski-patrol training session. It
was a beautiful day, sunny, clear, and cool, and it was wonderful
to be back on the snow. We spent the day conducting drills with
avalanche beacons, the signal transmitters that skiers wear to help
rescuers find them if they get buried by a slide. One group would
go out and bury a set of beacons, and others would then practice
locating them and digging them up. There were other exercises,
too, including working with pulley systems, rescue sleds, and
medical gear. On Sunday, we went off for a long ski on our own,
and when we got back to the cabin that night we were both
exhausted from our weekend's exertions.
   I spent an hour or so returning phone calls and checking in
with friends, hearing several times that David Bank, the San Jose
Mercury reporter, was still pursuing the theory that I had staged
the break-in as publicity stunt. I also heard that Bank had dined
with John Gilmore on the same evening that Julia had flown to
meet me in Palm Springs, and I began to wonder what John had
told the reporter. Certainly, the nature of the attack on Ariel indi-
cated a sophisticated perpetrator familiar with TCP/IP and Unix.
And those initial probes had, after all, come from toad.com.
   As I mused aloud about my suspicions as we ate, Julia and I
concluded that no matter how angry John might be with either
of us, he was too principled a crusader for electronic privacy to
have anything to do with a computer break-in. Still, the rumors
and the continued questioning of my motives made me curious
about where this train of events would lead, and we decided to
write up a chronological record of events to help us understand
and explain what had happened.
   Around 11 P.M. Andrew called. We had been in touch briefly
several times since Wednesday, but this was my first opportunity
for a full update. "What have you found so far?" I asked him, still
hoping that he could handle things at the Well on his own.
   "Tsutomu, I think you should come down here and help me,"
he said. ''I'm in over my head."
144                        BREAK-IN
   After having spent several days wading through the Well
bureaucracy, Andrew had finally begun to examine the stolen
software stashed on the Well system on Sunday morning, and it
turned out to contain much more than just my collection of files.
He had spent the day creating an inventory of the stolen materi-
al, growing increasingly alarmed by the value and sheer volume
of the contraband. "It's clear we're dealing with something other
than your average high school or college system cracker," he said.
   The software was located in a number of illicit accounts on the
Well, and my stolen files, so carefully deleted a week earlier, had
shown up again, in a different account. This fact alone suggested
that whoever had commandeered the Well's computers was cocky
enough to believe he could come and go, moving things around
with impunity. Andrew began rattling off a list of what he'd
found, but I interrupted, saying I wanted to go through it sys-
tematically. Rich Ress of the FBI had told me during that apolo-
getic phone call back in January that the Bureau assigned priori-
ty to its cases by dollar value. I decided that if they wanted dol-
lars, I would now give them dollars.
   As Andrew began once again to itemize his findings, I started
to respond by estimating the worth of each stolen program,
whether in market value or in development costs. Besides my
software, there was cellular phone software from Qualcomm, a
San Diego technology firm; lots of programs from an East Coast
software house called Intermetrics, which makes software devel-
opment tools; Silicon Graphics source code, the 3-D workstation
software used in creating most of Hollywood's movie special
effects; plus the computer security software that was supposed to
keep the company's programs from being stolen; sniffer logs from
Motorola's Semiconductor Products Sector Internet gateway
computer, which had captured information flowing into
Motorola's network, including passwords; various stolen pass-
words captured by other sniffer programs; an entire password file
from apple. com, Apple Computer's gateway to the Internet; and
sundry software tools for breaking into computers in various
   At the end of my tally the dollar count reached millions of dol-
lars in software development costs, a figure that didn't include
perhaps the most remarkable trophy, whose potential value I
                        KOBALL'S FIND                         I 45
couldn't begin to estimate: a long data file named 0108.gz, con-
taining more than twenty-thousand credit card account numbers
of the subscribers to Netcom On-Line Communications Services
Inc., an Internet service provider based in San Jose, California.
Many online network companies ask their subscribers to provide
full credit card information when they establish their accounts,
though such information is generally not stored on a computer
connected to the Internet.
   Credit card information wasn't Netcom's only loss. The thief
had also pilfered its subscriber password file, another parcel of
information that ordinarily should not have been accessible. The
version of Sun Microsystem's operating system installed at
Netcom takes some measures to protect that file: the passwords
in the file are encrypted, which theoretically makes them useless
to anyone who might stumble upon them, and the file contain-
ing the passwords is inaccessible except to someone who has root
access to their computers.
   Still, possessing a copy of that file would enable a thief to
unscramble some poorly chosen passwords. The method of
encryption used to encode them is well known. Given this, it
would simply be a matter of using that method to encrypt each
word in a large dictionary, then comparing the scrambled dictio-
nary words with the encrypted passwords in the file. For any
matches, the cracker could backtrack to the unencrypted dictio-
nary word and presto! a valid password. A break-in artist can use
a computer to run this kind of codebreaking attack quickly and
successfully against people who are careless enough to use ordi-
nary words as passwords.
   Andrew then turned to another category of stolen goods: e-
mail. Besides the contents of my mailbox, the thief or thieves had
rifled two other people's mail. Andrew and I both recognized the
name of Eric Allman, who was the author of sendmail, the stan-
dard Internet mail program. My guess was that Allman's mail had
been scoured for reports of new security flaws in sendmail, but
Andrew, sensitive to privacy issues, had not read the stolen mail.
The other name, which neither of us knew, was that of a Stanford
student named Paul Kocher, whose e-mail the thief had squir-
reled away for reasons unknown to us.
   The Well's monitoring had narrowed the intruder's comings
146                         BREAK-IN
and goings down to a forged account called dono, using the pass-
word "fucknmc"-which in itself sounded like a clue. In the
Unix and Internet world there is a grand tradition of using the
initials of your first, middle and last names as a log-in name, and
we wondered who was the "nmc" against whom the thief seemed
to bear a grudge.
   Andrew and the Well staffers on the case could see only what
went on locally when the thief connected from a remote com-
puter, but by tracing dono's activities on the Well, it was possible
to see a definite pattern of activity emerging and to make some
reasonable guesses about what he was doing elsewhere. Because
the burglar tools and stolen goods constituted evidence that no
clear-thinking criminal would want to leave lying around on hard
disks at home, the dono cracker was apparently using the Well as
an electronic storage locker. For each sortie against a computer on
the Internet, dono would fetch copies of his tools from the Well
in a predictable sequence.
   First he would come and take a garden-variety break-in pro-
gram that enabled him to get root on a less-than-vigilant com-
puter system somewhere on the Internet. Then, a short time later,
he would come back and pick up a "cloaking" program, which
would hide his presence on the raided system, at least from the
casual observer, by deleting traces of his activities from the system
logs. After finishing his nefarious tasks, the intruder would return
to the Well to pick up a sniffer program, which he could leave
behind at the raided site for gathering passwords that might make
it possible to break into other machines later.
   This was a very methodical criminal.
   But Andrew had also seen that the cracker, true to the form dis-
played in our San Diego break-in, was being sloppy. Once he had
run his cloaking program on a particular computer he'd broken
into, which removed evidence of his presence from accounting
files, he didn't bother to cover his tracks as he proceeded to carry
away his stolen files. Records of the actual file transfer, for exam-
ple, might be left elsewhere on the system. A casual observer
might not notice his activities, but anyone looking for such
behavior could probably track it easily.
   It took three-quarters of an hour for Andrew to describe all that
he had observed. "This is huge, Tsutomu," Andrew concluded.
                         KOBALL'S FIND                         I 47
 "We don't even have enough computers to provide adequate
 monitoring." Worse, he continued, even though he had caved in
 and signed all the nondisclosure documents the Well had placed
 before him, his activities were still being narrowly circumscribed
 by a woman named Claudia Stroud, the administrative assistant
 to Bruce Katz, the Well's owner.
   "I've been here a week, I've collected data," Andrew said. "I've
 got some idea of where these guys are coming from, but now I'm
 in over my head. Now it's your turn."
   He told me that Well executives wanted to have a meeting the
 next day with a small group of people to discuss how to respond
 to the attacks. I assured him that I would be present, but asked
 him if he could get it postponed until the evening; I wanted to
 get in one more day of skiing before leaving the mountains.
   In the meantime, I told Andrew he should gather some more
 information. I wanted data on the time and date of each file's
 access, and asked him to make a timeline of the cracker's con-
 nections to the Well. I also suggested that he make a more thor-
 ough search for backdoors and Trojans. It was important that he
 and the Well's monitoring squad not prematurely narrow the
 scope of their surveillance. We didn't want to be like the drunk in
 the classic joke, who upon losing his keys looked for them only
 under the streetlight because "that's where the light's best." He
 had brought one of the RDI computers with him from San
 Diego and I asked him to pick up the second one I had lent to
.Soeren Christensen, a friend who worked at Sun, once he arrived
 in the Bay Area. I was bringing a third machine to give us enough
 resources to both monitor and analyze the data.
   After we hung up I sat on the floor of the cabin and for many
 quiet minutes watched the fire dancing through the glass of the
 potbellied stove. Despite his protestations that he was in over his
 head, Andrew had uncovered a great deal of evidence. Real crimes
 were being committed, with no sign of letup, and there was now
 a hot trail to follow.
On Monday Julia and I woke midmorning.
   I sat up in the low bed in the center of the A-frame's loft and
stretched. The loft filled the back half of the cabin, and I could
see nothing but gray light filtering through the curtains that cov-
ered the picture window at its front. I thought again about
Andrew's call for help the previous evening. There clearly was a
case to pursue, and I could no longer in good conscience hope
that someone else would take care of it. That's why, after talking
to Andrew, I'd stayed up late dealing with voice mail and e-
mail-a ritual to prepare for reimmersing myself in the outside
world. There was yet another threatening message, and I for-
warded it to Andrew and attached a message asking him to con-
tact the authorities to trace the call.
   Even though it didn't look like a great day to ski, I was deter-
mined to get out once more. A light snow was falling, whipped
by a gusty wind until it blew sideways, and occasionally even
straight up into the air.
   I didn't have a particular plan, but it seemed logical to spend
a couple of days at the Well to size up the problem. There were
growing reasons to think that our intruder might be Kevin
Mitnick, not only the Oki and Qualcomm cellular source code
that had been stashed at the Well, but also other random indi-
cations- including the tip from the cracker Justin Petersen. But
I 52                         PURSUIT
there were also still reasons to rule out Mitnick as well, particu-
larly the sophistication of the IP-spoofing attack and the voice-
mail taunts, which I was fairly certain were not the work of a sin-
gle person (although Mitnick, of course, could have been part of
a conspiracy). If it was actually Kevin Mitnick, there would cer-
tainly be a lot of interest from law enforcement types. Levord
Burns, the field agent with whom Andrew had been talking,
worked for Rich Ress in the FBI's computer crime squad in
Washington, D.C.. The Bureau gave every appearance that they
were going to be helpful this time, but ever since I'd talked with
the FBI about computer crime at their Quantico training center,
my impression was that even if they had decided to be helpful, it
was no guarantee of success. I have respect for the integrity of the
agents I've met, but even they admit that they are usually out-
gunned when it comes to computer crime. The average field
agent has usually taken a training class so that he knows how to
recognize a computer at the crime scene, but he probably won't
know how to turn it on.
   I had no doubt, on the other hand, that the Bureau does
understand a great deal about the psychology of the habitual
criminal. At Quantico I also learned about techniques for track-
ing and catching serial killers, and how easy it is to get away with
serial homicide. FBI serial crime experts believe that there are
similarities between computer crime and other more violent
kinds of elaborately repeated crimes. It's a controversial idea, but
the FBI experts argue that the same compulsive behavior, and the
same craving for power drives both kinds of criminals. These
behavioral scientists theorize that in each of the cases the crimi-
nals have a need for a fix, which becomes increasingly frequent.
More relevant for my work is their assumption that in either type
of serial crime, the major investigative problem is information
management-organizing and marshaling the accumulated facts.
Frequently, they say, when they backtrack in a serial crime case,
they discover they already had the solution to the case much ear-
lier but didn't recognize it.
   Before I left the cabin I asked Julia if she wanted to come along,
even though I had no idea what might develop or where the trail
would lead. I thought that her organizational skills might be use-
                            BOTANY                            I 53
ful in our investigation. She said that she didn't know much
about computer security issues, and it was a chance to learn more
about them. She decided to come to Sausalito, but took a final
look at the poor conditions outside and opted to skip the ski.
   I finally headed over to the Tahoe Donner cross-country cen-
ter, around noon. Few people were on the trails, and the groom-
ing machines had left the course fast. It was well into the after-
noon, and the light was already beginning to fade when I called
Julia to pick me up. I didn't have time to change from my lycra
cross-country clothes before we headed for the Bay Area in her
   The weather was turning worse, and chain control was in place
on Interstate 80, making things even slower on our way out of
the mountains. It was merely raining by the time we got closer to
San Francisco and took the back roads north of the Bay to Marin
County. Around 8:30 P.M. we arrived at the Buckeye Roadhouse,
a trendy restaurant in Mill Valley, near Sausalito, where the Well
is located. We had planned a dinner there as an opportunity to
meet several of the Well's board members and other friends of the
Well in order to reach a consensus about how the on-line service
was going to handle the attacks. Andrew, who was staying in a
spare room in Peis home, had already arrived and before we sat
down he brought me up to date. More cellular telephone soft-
ware and a variety of commercial programs were discovered
stashed in other places on the Well. Today he had found the soft-
ware for a Motorola cellular telephone.
   He also told me of an odd but intriguing discovery. One of the
things he had come up with was a peculiar backdoor communi-
cation channel the intruder was using on the Well. Because the
attacker could become root on the Well whenever he wanted, he
could freely examine any other user's electronic mail. The mon-
itoring team watched him check a number of mailboxes, includ-
ing that ofJon Littman, the Marin County freelance writer, who
had a legitimate Well account. Littman was working on a book
detailing the exploits of Kevin Poulsen, the Bay Area computer
cracker whom Justin Petersen had helped implicate in the radio
station scam and who was still in prison charged with espionage
for possession of classified military computer tapes. He had also
I 54                        PURSUIT
been assigned a piece about Kevin Mitnick the year before by
  While Andrew had been monitoring the network, he had
watched as the intruder became root and copied a file from a
remote computer, a letter written by Kevin Ziese, an officer in
charge of the Air Force Information Warfare Center in San
Antonio, Texas. The intruder then logged in as Littman and,
withinLittman's own account, began to compose a message that
he addressed to the writer, with a note in the Subject line: "Here
you go :-) A vision from God." He then attempted to copy the
Kevin Ziese file into the Littman message, but got stuck when
he apparently could not figure out how to use the Well's mail-
editing software.
   He then abandoned the mail program and instead returned to
the Well as root once again and simply added the Ziese file to
Littman's mailbox file. The Ziese letter contained a long discus-
sion of the dangers inherent in the IP-spoofing attack, and it
referred to a conversation that Ziese had with me at the CMAD
conference. At the bottom of the letter, above Kevin Ziese's sig-
nature line, the intruder inserted a single line which said "****
Hey john [sic], Kevin is a good name :-)".
  Andrew was certain that this was the final clue pointing to
Kevin Mitnick. I was still hesitant to jump to that conclusion.
After all, there is no shortage of Kevins in the world. But clearly,
we both wondered the same thing: was Littman aware of this pri-
vate channel the intruder had created? There was no evidence
that he knew, or, if he had noticed, that it was anything more
than a taunt by the intruder, not a sign of complicity.
   We went inside to meet Bruce Katz and the Well board. A
large table had been reserved for us in the back. Bruce Koball,
the Berkeley programmer, had been invited, as well as several
other longtime members of the Well community. An old friend
of Julia and John Gilmore, Koball gave her a quizzical look
when she showed up with me. But there was such a din in the
Buckeye you could barely hear the person next to you, so there
was little chance for Julia to quietly explain what had hap-
pened. In any case, I was preoccupied by my dissatisfaction
with this noisy meeting spot, which was hardly conducive to
holding a confidential group discussion. It seemed we were vio-
                             BOTANY                            I 55
lating basic operational security principles at the outset of the
   The Buckeye was Bavarian in an antlers-on-the-wall sort of
way. I ordered salmon and Julia a shepherd's pie, while Andrew
opted for some remarkably large slab of meat, obviously taking
advantage of the menu.
   Bruce Katz sat next to me. An entrepreneur in his mid-forties
who had founded and run the Rockport shoe company before
buying the Well, he had thinning long hair and informal dress
that made him look more like a veteran of the 1960s than a
businessman. Over the din, I tried to brief him on what we'd
found. As if to underscore the urgency of the situation, shortly
after we sat down a Well employee who was monitoring the sys-
tem back at the office called to tell Pei that the intruder had just
used the Well as a jumping-off point to break in to Internex,
another commercial Internet service based in Menlo Park,
   I knew Bob Berger, the computer engineer who was Internex's
founder, because he had provided ISDN Internet connections for
Sun Microsystems at various times. I also knew something else
about Internex: it was Markoff's Internet electronic mail
provider. It occurred to me that this might be the reason for the
attack, but I decided not to voice this suspicion until I could
investigate. When Andrew and Pei decided they would call
Internex in the morning, I tried to persuade them that the com-
pany should be alerted immediately. But, nobody seemed anxious
to give up dinner and track down an Internex system adminis-
trator, who would probably be hard to locate at this hour, any-
    Katz wanted to know whether the intruder was a person who
could potentially damage the Well's system. Since we still
weren't sure who it was, we didn't have a good answer, and since
we had no idea how he might respond if he detected our mon-
 itoring, I told Katz the possibility of retaliation could not be
ruled out.
    The damage question cut to the core of the issue. It's a com-
 mon argument in the computer underground to say that break-
 ing into computer systems is morally defensible because all the
 trespassers are doing is looking, not tampering. Crackers also
I 56                        PURSUIT
like to claim that they are in effect helping make systems more
secure by revealing vulnerabilities to system operators.
   To me these are ridiculous claims. Once upon a time such
behavior might have been defensible, when computer networks
were research systems used only by engineers and research schol-
ars, although not many engineers or professors I know would
agree. In any case, today, when companies and individuals are
using computer networks as essential elements of their businesses
and lives, the cracker rationale is tantamount to my saying that it
would be permissible for me to break into your house and walk
around as long as I didn't take anything. Even if material like pro-
totype cellular telephone software isn't stolen, only copied, it
remains intellectual property that could easily give an industry
rival an important competitive market edge. In those cases in
which real damage is done to software and even hardware when a
cracker breaks in, companies are forced to spend tens of thousands
of dollars cleaning up. In a particularly complex computer it often
takes a great deal of effort just to figure out what has been dam-
aged, or what has been taken. There isn't any way to justify
Internet joyriding, and its most long-term and corrosive aspect is
that it causes network users to put up stronger barriers, destroying
the community spirit that has long been the hallmark of the Net.
   The conversation shifted to computer hacker and cracker cul-
tures, and how the potential for mischief was actually a great deal
more significant than what we tended to see. I pointed out that
among computer criminals, it is usually only the dumb ones who
get caught. Katz didn't seem comfortable with this line of rea-
soning, for he really wanted to believe what we were seeing was a
harmless prank. But to me, the Well's intruder didn't seem harm-
less. I explained to Katz how password-sniffing worked and how
it had the potential of permitting an electronic trespasser to gain
access not only to a single system, but to systems all over the
Internet. I also tried to explain that the only real security lay in
the extensive use of cryptography. The problem is that most cur-
rent crypto systems make networks more difficult and expensive
to use, and so people tend to avoid adopting them.
   It was clear Katz wanted to do the right thing, and he was will-
ing to learn about computer security. The problem was that
because he didn't understand the technical details as he himself
                             BOTANY                            I 57
admitted, he wasn't sure what the right thing was. The WelI runs
some private conferences used by consulting groups and other
private organizations. He wanted to know if it was possible at
least to fence those conferences off, and guarantee their security.
  Unfortunately it wasn't, I told him. I mentioned using digital
tokens, credit card-sized devices that produce a new password
every minute, but when I told him their price, he acknowledged
they were out of the question.
   "Couldn't we just shut the guy out?" Katz asked. He wanted to
know if it might not be sufficient to simply have all eleven thou-
sand Well users change their passwords.
   "Probably not with any degree of confidence," I answered. At
this point, since the interloper had been root for an unknown
period of time, at least many months, before he was discovered,
the WelI had to assume that alI of their operating system software
had been thoroughly compromised. In addition, there was no
way of knowing for certain that all of the accounts the cracker
had created had been identified. Maybe he'd stashed a handful of
accounts, and left them secretly sitting in reserve just in case he
was detected and needed to use them later. Even worse, we were
pretty sure he could make himself root from a normal account.
   "If you try to shut your doors by changing passwords and clos-
ing his accounts, he's almost certain to have hidden a Trojaned
program somewhere that would allow him to come right back
in," I said. "Only this time you won't know where he is."
   I summarized the list of stolen software that Andrew and the
others had found, and admitted, "I stilI don't know exactly what's
going on, but one thing I do know is that there is a huge amount
of data of great commercial value that someone is hiding there."
   I began to understand that the Well directors were looking for
easy solutions and assurances that I couldn't give them, because I
hadn't even visited the Well yet. This was something I might not
be able to accomplish on my own, I warned. I would probably
also require support from other Internet service providers, as well
as law enforcement officials. Gathering data on the WelI was a
start, but it was possible I would eventually need trap-and-trace
information from telephone companies to actually locate the
intruder. I tried to convey my sense of urgency as I explained all
this, and told them that in a situation like this you had to go full
I 58                        PURSUIT
speed ahead, or you might as well forget about it. One hint
leaked to our intruder that we were watching him, and any trail
might instantly be obliterated. The main issue confronting us was
whether the Well's directors were willing to keep their system
open, and not do anything to tip off the intruder that he had been
detected, so that we might have a chance to track him down.
   The Well group listened carefully, but it was clear they were in
a high state of anxiety about how their users might react both to
the break-ins and to management's reponse. The Well has always
been an unusual place in cyberspace. Besides attracting a coterie
of Bay Area hackers and Deadheads, the Well had also become a
favorite online watering hole for the computer media digerati-
technology writers who engage in online gossip and who were
likely to be among the most verbal critics of any missteps by Well
management. From what I had heard, the Well as a community
has its own strongly held sense of values, and anyone who trans-
gresses the group's conventions does so at the risk of becoming a
social pariah. As a recent arrival to this digital world, Bruce Katz
could not afford to be shunned.
   The Well's vice president of administration, Claudia Stroud,
who had been Katz's top lieutenant before he purchased the
Well, was nervous about the liability the company might face
because of our surveillance operation. Besides the matter of an
intruder who was reading other people's mail, she noted that
there were privacy-rights activists among the Well's members
who would freak out when they learned that investigators had
been systematically filtering all of the system's data traffic on the
   "How will keeping the doors open longer and not telling the
users what's going on yield any better results?" she demanded to
   Claudia, who related to Katz with the protective familiarity of
a big sister, commingled with time-tested respect for a mentor,
was probably only doing her job. But from my perspective, the
only sure way for the Well to return to normal would be if we
found the intruder, and it looked like Claudia might try to stand
III our way.
   "From what I can see," she said, "the Well has been in sus-
pended animation for the past week and a half, and this investi-
                             BOTANY                           159
gationhas little to show for it."
  The Well had been planning to transfer its operations to a new
Sun Microsystems SPARCcenter 1000 computer, and through-
out dinner the discussion kept returning to the question of how
quickly they could and should cut over to the new equipment.
Replacing all of their hardware and software might temporarily
improve their security situation, but it would complicate our
      .,             .
mOnItonng operation.
  By the end of the evening, Katz was sobered by the extent of
the break-in and the sheer quantity of software, credit card
information, and data files we'd found. He seemed to have
decided that the only way to make the Well secure was to shut it
down, and transfer its operation to a new computer with trust-
ed software. And yet, we seemed to have made our point that the
best way to ensure security would be to put this cracker out of
  ''I'll give you a little more time," Katz said at last.

After dinner we followed Pei and Andrew in the light mist that
had replaced the day's rain to the Holiday Inn nearby in San
Rafael, where Julia and I were going to stay. Andrew was driving
the red Cherokee the Well had rented him, which he had taken
to calling the +4 Jeep of Intimidation, a reference to the power-
ful imaginary weapons players are awarded in fantasy and role-
playing games like Dungeons and Dragons.
  As I followed the jeep, I thought to myself: I wonder whytheWell
thought our investigation required a four-wheel drive vehicle?
My second thought was: It's going to be a pain topark, but at least
we can park on top ofthings.
   Levord Burns, the FBI agent, had asked Andrew to call him
after the meeting to tell him what the Well had decided to do. So
even though it was midnight when we reached the hotel, and 3
A.M. in Virginia where Burns lived, I telephoned. He sounded
sleepy, but calls in the middle of the night are part of life for a
Bureau field agent, and within a few moments he was speaking
in the formal, somber demeanor that we were familiar with.
   I reviewed what the monitoring had turned up so far, and told
him that I was going to the Well's operation center the next day
160                          PURSU1T
to look at the data. While we talked he told me that, despite hav-
ing recently been made the Bureau's primary field agent for com-
puter crimes, he had little in the way of a technology background
or experience in cases that involved the theft of information.
   "I usually deal with bank robberies, Tsutomu,' he said.
   I concluded by telling him the Well had agreed to let us con-
tinue monitoring for a while, and he said he would wait to see
where it led.
  Andrew and Pei left after the phone call. Before Julia and I fell
asleep, she said that she no longer felt like she had a home, that in
recent weeks home seemed to have become whatever hotel we
were currently staying in. ''And as hotels go," she said, "this one is
a definite step down from the resorts we've been staying at lately."

We arrived at the Well around eleven-thirty Tuesday morning.
The nondescript office building was nothing like the nearby
quarters, backing up to a row of houseboats, in a funky section
of Sausalito where the online service began in 1987.
  The original site was the offices of the Whole Earth Review, and
the Well-an acronym for Whole Earth 'Lectronic Link-was
closely associated with Stewart Brand, one of Ken Kesey's Merry
Pranksters and the creator of the Review, and of the Whole Earth
Catalog. Originally a leading light of the back-to-the-land move-
ment in the 1960s, Brand had written an article for Rolling Stone
magazine in 1972 in which he described a crazy group of
researchers at the Xerox Palo Alto Research Center who were try-
ing to reinvent computing. Within a few years they had succeed-
ed, coming up with the forerunner of the personal computer.
  In the late 1970s, when the personal computer industry first
emerged, it was still largely a collection of hobbyists with a strong
countercultural flavor. In the late 1980s the Well mirrored this
same eclectic mix of hackers and hippies. Well members began
connecting first from around the Bay Area, and later from all over
the country, to chat about things that were on their minds. When
the Information Highway hype took off, dozens of reporters
wrote stories about the Well, giving it influence far out of pro-
portion to the size of its membership. And so it had a certain
cachet in 1994 when Katz, already a Well investor, bought the
                             BOTANY                            I 6I
remainder from the nonprofit group that controlled it and
embarked on an ambitious plan to turn the Well into a signifi-
cant national for-profit service.
   One of his first moves was to shift the Well from its houseboat
neighborhood to an office complex several blocks away where we
arrived Tuesday morning. Pei led Julia and me through a large
open space where the support crew and administrative staff
worked on PCs and Macs, and showed us to the back, where her
office and the computer systems and file servers were located.
Down the hallway was a large open closet with a rack of modems
so that users could dial in and connect to the Well.
   I could tell that Pei, a woman roughly the same age as Julia, was
competent, bright, and capable, but she seemed a little too hesi-
tant. The job had been a one-person operation when she began
at the Well in the middle of 1994, but it had quickly grown to
the point where she oversaw four or five people, and she was
clearly new and unsure of herself about the management side of
her job. She complained to us about how hard it was to get the
Well to listen to her, particularly about security matters. It had
taken Andrew coming in as an outside expert to recommend
things, and to get her the support she needed.
   Julia and I had shown up in time for lunch, actually our
breakfast, which was being brought in for the small group of
systems people that ran the Well under Pei's guidance. In order
not to tip off anyone at the Well who didn't have a need to
know about our activities, we were to remain hidden in the
small room at the back of the building where Andrew had been
operating for the last week. When we walked in Andrew was
faxing a page of information to Levord Burns that the trap and
trace at UCSD had turned up about the voicemail messages
that had been left for me. That same morning Andrew had
learned that the voicemail messages had come in over Sprint's
long distance lines; that probably meant our caller wasn't in San
Diego. Andrew was sending the information to Burns in hopes
that the FBI would be able to get precise location information
from the phone company.
   Andrew had also been following up on his suspicions that we
were up against Kevin Mitnick. Earlier, Andrew had called the
local FBI office, which referred him to Kathleen Carson, an
I 62                        PURSUIT
agent in Los Angeles, who was apparently running the Mitnick
investigation for the Bureau. She had been willing to tell him
only a few things, and things that weren't very useful, including
some of the names of a number of Mitnick's past and present
associates, including Kevin Poulsen, Justin Petersen, Eric Heinz,
Lenny DeCicco, Ron Austin, and Lewis Depayne. She said the
FBI knew of one recent computer account Mitnick had used,
named "marty," but she wouldn't reveal any of the specific
Internet sites that were involved. When Andrew ran off the sites
we knew were related to the Well's break-in, she merely grunted
a few times.
   As we ate the mild Chinese scallops, shrimp and snow peas,
and kung pao chicken that had been brought in, I began to assess
what we now knew. Since the previous night the Well had seen
more traffic to and from Internex, so I called Bob Berger and left
a message warning him that Internex had been broken into. Then
I called Markoff and alerted him that somebody might be read-
ing his electronic mail. He said that more than a year earlier a pri-
vate message sent to him at the Well had shown up on a public
newsgroup, so he had largely stopped using the Well for mail and
instead had set his Well account to forward his messages to his
New York Times account which was handled by Internex. Now
Internex did not look very secure either.
   Because the Internet had become an essential tool for most
technology reporters, and every journalist dreads getting
scooped, Markoff was naturally concerned about someone peer-
ing over his shoulder and reading his mail. But he agreed to do
nothing to tip off any snoops and wait to see what my investiga-
tion turned up. He did, however, take one precaution. His com-
puter at the Times bureau in downtown San Francisco automati-
cally connected to Internex every hour to check for new mail. He
decided he would increase the frequency to once every twenty
 minutes so that mail waiting at Internex would be vulnerable to
snooping for a shorter period of time.
   After I finished my calls I turned my attention to the Well's
 monitoring operation, which was obviously in disarray. Pei was
 collecting some information on a Sun workstation using a stan-
 dard sniffer program called "snoop," while Andrew was collecting
 different data on an RDI laptop, which he had connected to the
                              BOTANY                             I 63
WeII's internal network. I was disturbed by this setup, because it
provided no easy way to compare the findings from Pei's and
Andrew's machines. Worse, nobody seemed to be doing much to
analyze the data they were coIIecting.
   Some intriguing information, however, had already shown up.
In addition to the Computers, Freedom, and Privacy account,
and the dono account that Andrew had been watching, there
were at least four others being used by the intruder: "fool," "fair-
demo," "nascorn," and "marty"-yet another indication that it
might be Mitnick. All of the accounts were demo accounts, so
there were no biIIing records kept for them. This suggested that
whoever was behind the break-ins had some detailed knowledge
of the WeII's accounting practices, and had set up or taken over
accounts where a bill wouldn't tip off the account holder that
someone was racking up unauthorized charges.
   Pei and Andrew had also generated a list of other Internet sites
which they now knew the intruder was either coming from or
going to over the Internet. This included Internex; Colorado
SuperNet, a commercial Internet service with headquarters in
Boulder; Motorola Corporation; NandoNet, the online service of
the Raleigh News and Observer, and Intermetrics. There were also
connections from a New York City-based public access Unix sys-
tem which seemed to have a suspicious name: escape.com.
   Also, there was a list of comings and goings from Netcom
whose customer credit card numbers had been stashed on the
Well. The day before, Andrew had caIIed Netcom and let them
know that one or more intruders had been ransacking their sys-
   As Pei and Andrew spoke of their efforts, it seemed to me that
they had prematurely narrowed their scope of inquiry. They
seemed to be saying, "We're looking at these five stolen accounts,
and we're looking at the usage of these accounts."
   It was an attitude I had feared the night before, when I talked
to Andrew from Truckee.
   "How do you know this is all there is?" I asked him. It was
obvious we needed to cast a broader net.
   Andrew had many pieces of paper stapled together. Some had
listings oflog-in and file-access times, but it was really hard for me
to teII what they were. Nothing had reaIIy been sorted into any
I 64                        PURSUIT
rational order that I could see. If we were going to catch the
intruder, we needed to systematically conduct what the intelli-
gence community calls traffic analysis. Rather than looking at
what was in each individual connection, I was more interested in
seeing when the connections took place, where they came from or
went to, and what else happened simultaneously. And before we
could find our way to the bigger picture, I was going to need to
understand how the Well's internal network was laid out, and find
a single point where we could see all the information coming from
and heading toward the Internet. Unfortunately, these steps
would have to wait. The Well had scheduled a meeting at 2 P.M.
with a Justice Department attorney and the FBI to discuss the
break-ins and the stolen software. I was to serve as the designated
technical expert.

The gathering took place at the offices of the Rosewood Stone
Group, Katz's holding company located just a couple of blocks
from the Well. Julia, Andrew, and I attended, along with Pei,
Claudia, and the Well's lawyer, John Mendez. Representing the
Government were Kent Walker, an assistant U.S. attorney in San
Francisco, and two FBI agents from local field offices: Pat Murphy
from San Francisco and Barry Hatfield from San Rafael. I'd heard
ofWalker, who had formerly been involved in Justice Department
computer-crime policy and cryptography issues in Washington.
He had a reputation for being tough on computer-crime issues but
I didn't have any idea how technically savvyhe was. Now that I was
meeting him, an athletic six-footer in his early thirties, he struck
me as having a quick mind and aggressive demeanor.
   Andrew and Pei began describing some of the keystroke mon-
itoring results captured in the past week, and talked about ana-
lyzing the intruder's behavior patterns, like trying to study a spec-
imen. As I listened I became increasingly impatient. Like partic-
ipants at academic computer security conferences focusing on
theoretical results rather than real events, they were more con-
cerned with classifications than direct action.
   "That's fine and well, but it's all botany!" I interrupted, no
longer able to contain myself "We're searching for a carbon-
based life form!"
                             BOTANY                           165
  The room was silent for a moment, but my outburst had the
effect of refocusing discussion away from what we could do to
secure the Well from the threat, toward my point of view, which
was that the only way we could secure the Well against the threat
was to go and apprehend it. Instead of a duck-and-cover posture,
we needed to shift into attack mode.
  I began outlining a plan to establish a base of operations at the
Well, and then to quickly move in whatever direction our sur-
veillance operations led us. In practice, my plan was for an orga-
nization like that of a mountain-climbing expedition. We would
have an advance team and a base team. We would jump forward
through the network until we'd pinned the intruder down to a
specific location. And when we actually found him? I figured that
was the FBI's problem.
  In such situations, when I'm trying to guide the agenda, I tend
to talk very fast. I later learned I'd overwhelmed the FBI agents,
neither of whom had much technical expertise. "I didn't under-
stand a word he said," one of the agents told Walker later. "He
was speaking at 9600 baud, and I can only listen at 2400."
  To drive home my point that we were combating a live oppo-
nent, a carbon-based life form at the end of the wire, I used the
speaker phone to call my voice mail in San Diego. A new message
had been left there, one that had been sent during the previous
week and which I had listened to for the first time the day before.
It seemed that my antagonist was unhappy that I had turned
publicity's spotlight on him by putting his earlier messages up on
the Net as tweedledum and tweedledee.
   ''Ah, Tsutomu, my learned disciple," he began in a bogus Asian
accent, and then he started to sputter like someone who hadn't
rehearsed his lines perfectly: "I see you ... you put my voice for
Newsweek . . . you put it on Newsweek. And you put it on the net.
Don't you know that my Kung Fu is the best? My Kung Fu is
Great! Why you put my voice on the net?
   "This is not good. Have I not taught you, grasshopper? You
must learn from the master. I know ... I know all the techniques
and styles. I know tiger-claw style. I know train ... I know crane
technique. I know crazy monkey technique.
   ''And I also know rdist and Sendmail. And you put it on the
net. I'm very disappointed, my son."
I 66                       PURSUIT
   It was apparent that I had gotten my intruder's attention.
Which was what I'd wanted. He had risen to the bait, and with
the trap-and-trace data from the call, we might be able to begin
homing in on his location. Playing the message also reminded
everyone that we were after a real criminal, not just some cap-
tured lines of Unix commands.
   A short time later, we were interrupted by a conference call
from Netcom, the company that had managed to let a thief carry
off its customers' credit card information. There were three vice
presidents from the company on the other end of our speaker
phone, and they seemed very anxious to be cooperative, giving us
a number of contact names. I suspected from their tone they were
worried that somehow they might be held responsible for the
attacks on the Well, and they wanted to make it plain to the law
enforcement types gathered in the room that they were willing to
cooperate in the investigation. Walker and the FBI agents said
they would be back in touch.
   Earlier in the day when I had looked at the sheet showing log-
ins to one of the Well's hijacked accounts, I had immediately
recognized one of them-art.net, Lile's machine at Mark
Lotter's house. It was the same machine that Kevin Mitnick had
taken over the previous fall. More and more clues were point-
ing to Mitnick, both at the Well and in my break-in, and the
FBI agents, Murphy and Hatfield, began to go through their
files on him.
   Murphy said the Bureau had a lot of information but couldn't
share very much with us, only what was in the public record.
Trying to figure out what they could release, the agent excused
himself to call the Los Angeles FBI office. L.A. was reluctant to
release anything but allowed Murphy to go through the material
he had in his folder and read "sanitized" portions to us.
   While he looked through the file I wandered over and peeked
over his shoulder and saw a document stamped "Confidential"
and a Kevin Mitnick "Wanted" poster.
   He read aloud the list of sites suspected of break-in since
Mitnick had gone into hiding in late 1992: the Los Angeles office
of SunSoft, Sun Microsystem's software subsidiary; the University
of Southern California; Colorado SuperNet; Novatel, a cellular
                             BOTANY                           I 67
telephone manufacturer; Motorola; Pan American CelIular;
Netcom; Fujitsu; Qualcomm; Oki; US West; and L.A. CelIular.
   If they were right, Mitnick was certainly obsessed with celIular
   The FBI documents also described a raid in Seattle the previ-
ous faIl in which the target had narrowly avoided being cap-
tured. Without knowing who their suspect was at the time,
McCaw CelIular security officials, a private security investiga-
tor, and the Seattle police had conducted a surveillance opera-
tion to track someone who was making fraudulent celIular
phone calIs and using a computer and a modem. After follow-
ing their suspect for a number of days they went to his apart-
ment near the University of Washington, and when no one
answered they broke down his apartment door. The team con-
fiscated his equipment, which included a portable Toshiba
T4400 computer and a lot of celIular telephone gear, and left
him a John Doe warrant. The Seattle police staked out the
apartment after the search for several hours and then left. The
suspect, who they later determined was indeed Mitnick from
the data on his computer, came back to his apartment, spoke
briefly with his landlord, and vanished.
   Oops, I thought.
   Another item in the FBI documents concerned Mitnick's pos-
sible whereabouts. The Los Angeles FBI office had information
that, besides Seattle, he had been in Las Vegas, and more recent-
ly, in Boulder, where the Los Angeles agents believed he might
still be residing. In fact, the agents now told us, the Los Angeles
office apparently was working with the operators of Colorado
SuperNet in an attempt to monitor the intruder's activities and
was confident it was closing in on its quarry.
   Murphy asked me if it sounded plausible that Mitnick might
be running his computer modem through a celIular phone. It
didn't sound very likely, I replied. I had tried it, and the trans-
mission reliability was quite poor, for calIs tended to be dropped
repeatedly. Data transmission might have been plausible with a
 powerful three-watt phone, but with the O.6-watt hand-held
 units the FBI believed that Mitnick favored, it didn't seem very
 practical. It would take an enormous amount of patience,
I 68                         PURSUIT
because modems tend to deal poorly with the automatic hand-
offs that happen in the cellular telephone network as the phones
move from cell to cell.
   "If he is going over cellular he'll have an easily identifiable sig-
nature, because he'll have to constantly reconnect," I told them.
I made a mental note to look for any telltale signs of repeatedly
dropped connections that we might have collected in our net-
work traffic data.
   Finally the FBI agreed to share with us the accounts and pass-
words Mitnick had been using on other systems, including the
"marty" account. A password for one of these' accounts was
pw4nl. It occurred to us the most obvious translation of that
shorthand was "password for the Netherlands"-a country where
the computer underground was still extremely active, despite the
fact that the Dutch had finally enacted computer-crime laws.
From Andrew's monitoring we had already learned that the Well
intruder had an account on a Dutch machine called hacktic.nl,
frequented by crackers. It was operated by a Dutch anarchist
computer group known as Hacktic.
   I wasn't sure how much credence to give any of the FBI data,
since a lot of it was from the computer confiscated in Seattle,
which Mitnick knew was in their possession.
  Some discussion followed about whether Mitnick might be
violent and whether the Well was in any physical danger.
  "You know John Markoff wrote the book about Mitnick," I
said. "Why don't we call and ask him?"
  The FBI agents didn't think very much of the idea of patching a
newspaper reporter into the meeting, but Walker overruled them.
When we reached him on the speaker phone, Markoff explained
that everything he knew about Mitnick was already in Cyberpunk,
or in his front-page Times article the previous July. He also said that
he, too, was skeptical about whether Mitnick was the culprit. He
said he had heard the break-in was actually the work of a shadowy
group of people acting in concert, none of them Mitnick. But if it
was Mitnick, Markoff said, there was no reason to think Mitnick
had the capacity for violence. A story recounted in Cyberpunk
described how, in one of Mitnick's first arrests in the early 1980s, a
Los Angeles detective pulled him over on the freeway, and he began
                              BOTANY                            I 69
  After everyone was satisfied on that score I pressed Walker about
the legal limits of the surveillance operation we were planning. One
of the biggest privacy issues on the Internet concerns the rights and
responsibilities of commercial systems operators . .fu data packets
flow through their networks, system operators potentially can
record every keystroke and every piece of data, effectivelymonitor-
ing every action and every conversation. Packet sniffers, like the
ones we had set up at the Well, can be employed both responsibly
and irresponsibly. In setting up our filters at the Well we were
attempting to capture packets for only the sessions we were trying
to monitor. Often, it was difficult to draw clear boundaries. There
was no way of knowing if there was one or several intruders, and it
appeared that he or they were using a half-dozen or more separate
accounts. There was a good chance that some innocent data was
being caught in our wide nets. We had a brief run-through of the
provisions of the Computer Fraud and Abuse Act and the
Electronic Communications Privacy Act for guidelines as to what
we could and couldn't do in our investigation. The laws allow the
use of monitoring where fraud or crime is suspected. Walker and
the FBI agents said that what we were doing should be covered by
these laws.
   "This is a situation in which you're not going to be acting as
technical support for us," Walker said. "We're going to be serving
as legal and administrative backup for you." His attitude
impressed me. Up to this point I really hadn't held out much
hope we had a chance of finding the attacker, for I'd seen many
of these investigations botched before by the FBI.
   I told them I would need several STU-Ills, special scrambled
government telephones for secure communications. Kent said he
didn't know about the STU-Ills, but he had access to lots of
Clipper phones. They were based on the data scrambling chip
with the eavesdropping back door that the National Security
Agency had been trying to convince the government and public
to adopt-without notable success. I said I'd rather have the
   Finally Claudia and Mendez raised the Well's concern about
potential liability if they kept their system open while we con-
ducted surveillance. They asked if the Justice Department could
give them a letter endorsing their decision to continue operating
I 70                        PURSUIT
business as usual, and Walker agreed to provide them with such
a document.
   The meeting concluded at nearly four o'clock, and Julia and I
stayed behind in the conference room to return various calls I had
received on my cell phone during the meeting. One was from
Mark Seiden, a Unix hacker and a computer security expert who
had agreed to help Internex deal with their security problems.
When I'd come to the Well that morning Andrew had told me
that the monitoring team had seen the intruder move a 140-
megabyte file containing the contents of my home directory on
Ariel to Internex the night before, and I began to feel as if we
were dealing with a squirrel burying his nuts, scurrying around
and hiding them in various holes allover the Internet. When I
returned Seiden's call, I told him about the file and said I wanted
it deleted. Since we didn't want to tip off the intruder, however,
we agreed that Seiden would erase the file and then send the real
user of the account a message that said something like, "We've
deleted your file because you've exceeded your disk space alloca-
tion. We've told you time and time again not to leave huge files
lying around."
   After all the calls were dealt with, Julia and I walked back down
the street to Pei's office at the Well. Claudia had been hovering
around waiting to present me with the same document she had
forced Andrew to sign, a nondisclosure agreement intended to
keep me from mentioning anything I learned about the situation
to anyone outside the Well. It had already created a huge prob-
lem for Andrew, who had been trying to warn other companies
that their systems had been broken into, and software had been
   Because of the agreement he'd been reduced to calling people
and saying, "I can't tell you who I am, and I can't tell you the
details of what happened, but I want to let you know you have a
security problem." It was an impossible constraint to work with,
and earlier in the day I'd suggested he ignore that part of the
   Claudia was also trying to enforce her understanding that all
the stolen software found stashed on the Well was the Well's
property. This was creating another tremendous headache for
Andrew in attempting to talk to the victims of the theft and to
                             BOTANY                             171
get their assistance. I explained to her that another company's
intellectual property did not automatically become the Well's
property just because someone had stolen it and hidden it there.
She was worried that if it was disclosed that the Well was the stag-
ing area for break-ins around the Internet it would be liable for
any damages. I pointed out to her that the Well might have
equally serious liability problems if it became known that the ser-
vice was aware of break-ins at other sites and failed to notify the
victims, as her constraints on Andrew had already made happen
III some cases.
   Finally, I tried unsuccessfully to convince her that what should
matter most to the Well was lowering the roadblocks and letting
us go full speed ahead if we were going to have any chance of
solving their problem.
   "Tsutornu, I have to ask you to sign this to protect the Well's
potential liability in the investigation," she repeated.
   I stared at her trying to silently communicate, I have no inten-
tion ofsigning anything so ridiculous, but finally tact got the bet-
ter of me. "I don't think I can agree to this now, but I'll review it
and get back to you." What I actually meant was, Which part of
 "no" don't you understand?The fact I'd agreed to take a look at the
document appeared to mollify Claudia, and as I went back to
work I couldn't help thinking about what a friend had once told
me: diplomacy is the art of saying "nice doggy" until you can find
a stick.
   Having spent the better part of a day dealing with bureaucrats
I was at last able to turn my attention toward trying to under-
stand the topology of the Well's network. Andrew had attached
an RDI Powerl.ite to the network in the right place, so that all of
the Well's packets would flow past his computer, but odd things
were happening. It was soon apparent the Well's routing was
totally scrambled, so that more than a quarter of the packets in
its internal network were moving in an extremely inefficient and
round-about fashion. One of the router computers was throwing
up its hands and sending packets to another router to make deci-
sions on how to send each bundle of data to its correct address. I
felt a little like a plumber who shows up at a customer's house
and has to tell the owner that someone has routed the bathroom
pipes through the bedroom.
I 72                         PURSUIT
   The screwed-up routing wasn't my problem. What was impor-
tant was that we begin logging the right packets as quickly as pos-
sible. We wrote a variety of filters to capture packets both com-
ing into and going out of the Well. By putting together a watch-
list of all the known compromised sites and then logging other
suspect places data might come from, we had a good chance of
creating a complete record of the intruder's activities.
   What I had in mind was to start by putting a broad set of filters
in place on two separate computers to make certain we had redun-
dancy. I wanted to search a lot ofsessions for a little while to watch
for our intruder's telltale signature and then narrow the focus
again. That way we would be able to see if any covert activity was
being missed. As we set the system up, however, I realized the Well
was the busiest system I'd ever dealt with, and there was far too
much data if we monitored both directions, so I confined our
monitoring to inbound data. By ten o'clock, I thought I under-
stood what it would take to get the packet logging and filtering
systems set up, so Julia, Andrew, and I took off for dinner.

The three of us drove in the +4 Jeep of Intimidation to the
Cantina, a Mill Valley Mexican restaurant Julia knew. According
to house legend, Carlos Santana's father once played there in a
mariachi band.
   Over dinner we talked about the trouble Andrew had gotten in
for being up in northern California helping me. My deal with Sid
Karin had been that SDSC would contribute his salary for a few
weeks, but somehow word hadn't reached Andrew's managers. I
told Andrew that I'd placed a call to Sid about it earlier in the day,
and it looked like it would get straightened out. We also spent a lit-
tle time chatting about Andrew's academic career and his hunt for
a new Ph.D. thesis adviser. I told him that I would be glad to pro-
vide advice and direction, but he was going to need to find some-
one else to be his formal adviser and handle administrative issues.
   Sometime after 11 P.M. we drew up a list of things we needed
to do to get our monitoring systems completely in place, and
drove back to the Well. Pei had gone home at a sane hour, but
several people were still monitoring in the cramped back room
that was the Well's network operations center. We were capturing
                             BOTANY                            I 73
tens of megabytes of data each hour, far more than we could hold
on our disks, even overnight, so we tightened our filters.
   After midnight I began looking through the log file data we
had gathered during the previous day and immediately found
something: the keystrokes of the intruder looking through
Markoff's directory and mailbox. As I studied the data I could
easily see how he'd found his way to Internex: he had simply
looked at a file called forward in Markoff's Well directory that
automatically routed his electronic mail to Internex. I also saw
that he was looking through a number of mailboxes on the Well
besides Markoff's and Littman's. He had gone through the mail-
boxes ofEmmuel Goldstein, the editor of the phone phreak mag-
azine 2600, and who in real life was Eric Corley; Ron Austin, a
Southern California programmer who had been in trouble for a
number of computer crimes; and Chris Goggans, a reformed
member of the computer underground who published an online
computer underground magazine called Phrack.
   By 2 A.M. we'd done everything we could within reason. Julia
and I did not want to spend another night at the Holiday Inn, so
we drove across the Golden Gate Bridge into the city. We ended
up in the guest room at Dan Farmer's near the park. Earlier in the
day I'd called Dan and told him we'd found the source code to his
SATAN program and his electronic mail on the Well. I'd been
hoping to talk to him about the break-ins, but by the time we
arrived, he'd already headed out for the San Francisco club scene.
   I could brief him later. For now, I knew we had cast our nets as
widely as feasible. It was a question ofwaiting and seeing what we
were able to haul in. We had already seen suspicious traffic from
Colorado SuperNet, Intermetrics, and Netcom, and it looked
like we would soon have to make a decision about which way to
head upstream in the InterNet. Somebody had already estab-
lished the rules we were playing by, and now that I was joining
the game, I had decided to dive in and not look back.
We returned to the Well late the following morning. Any doubt
over whether I was committed to the hunt had been erased the
day before. I've always felt strongly that the way you do some-
thing matters as much as what you do, and if! was going to hunt
this thief! felt it was unacceptable to approach the challenge with
any less than all of the intensity and focus I could muster.
  There seemed to be a slowly accumulating body of evidence
that it was Kevin Mitnick who was sitting with a portable com-
puter systematically launching attacks throughout the Internet,
but there was still no conclusive proof. Was he directly responsi-
ble for stealing my software in December? The evidence was still
sketchy. What I did know from the data Andrew had collected
earlier was that even if Kevin himself hadn't been the one who
had attacked my machine, the Well intruder had in his posses-
sion a copy of my software within twelve hours of the original
   Now the chase was on, and the challenge was to move forward
faster than any leaks might get back to our intruder. Security had
become a real concern, because I realized people were talking
about me, and that my days of comfortable anonymity were end-
ing. Walking back to Pei's crowded office that morning, one of
the Well system-support people stopped me and said, "Didn't I
see your picture in the papers?" The publicity about the break-in
was obviously starting to complicate our activities, and it would
                        "YOU LAMERS!"                        I 75
be a disaster if somebody mentioned something about my pres-
ence on the system and it was picked up by the intruder.
   Pei went to the Well employee a little later, and asked if he
could please keep it quiet. I had the impression that both
Claudia and Pei thought they could keep things under wraps,
although I feared that was already proving to be impossible. In
fact, the situation almost immediately became worse when
Kevin Kelly, the editor of Wired and one of the founders of the
Well, showed up and asked if he could take my picture for an
article in his magazine.
   "It would be better if we did it tomorrow," I mumbled and
tried to disappear quickly.
   Slipping into the back room, I began to look over the progress
we had made in setting up our monitoring station. One of the
RDI machines was now logging data along with Pei's
SPARCstation, and we'd apparently done pretty well at tuning
our filters the night before, as the amount of data we were saving
for later analysis had become a little less overwhelming. However,
our software tools wouldn't read the data that was being filtered
on Pei's SPARCstation and it was being used for other tasks as
well, so we kept working with the second RDI trying to attach a
disk Pei had lent us.
   Tuesday had been a fairly quiet evening. Our intruder had only
made scattered appearances, giving Andrew the time to return
phone calls. He learned that the security people at Colorado
SuperNet had been detected by the intruder, who quickly delet-
ed all the files he'd stored there and then left them a cocky mes-
sage: ""\1: Iamers.I"
   I took note of the times at which the intruder was active. He'd
logged in to the Well at around 8 A.M. on Wednesday. If he was
keeping normal hacker hours, working till the wee hours and
sleeping late, it seemed obvious he wasn't in our time zone: the
Midwest or the East Coast would be more likely. The data also
seemed to suggest that our attacker was a single individual and
not a group, for there was never more than a single log-in session
using the stolen accounts at anyone time. Furthermore, just as
Andrew had described on Sunday night, his pattern seemed to be
 remarkably repetitive. The Well was clearly a staging ground or
jumping-off base from which, over and over again, he would
176                         PURSUIT
come and fetch his tools and then take them to the site of a new
   My growing suspicion that Kevin Mitnick was the Well's
intruder explained some things that had puzzled me earlier. I had
one of those strange flashes of insight that I occasionally get as I
was scanning through text on the workstation display and I saw
the letters VMS scroil by. VMS is DEC's operating system, and I
recalled that several years earlier I had read in Cyberpunk that
Neill Clift was the British computer researcher who specialized in
finding security loopholes in VMS. Neill Clift might have the
right middle initial to match the password "fucknmc." Might
Mitnick have some score to settle with him?
   I called Markoff to ask him to check on Neill Clift's middle ini-
tial and then invited him to come over and see our monitoring
operation in action. Markoff arrived about an hour later, and we
looked through some of the keystroke information we'd captured,
showing what the Well's intruder had been up to.
   Claudia also paid a visit and asked me what the status of our
operations was. The Well board was meeting the next day, she
said, and it would be making a decision about whether to take
the Well computers off-line or not.
   "We are feeling very exposed," she announced, "and I think we
should consider taking steps to resecure the system, like remov-
ing the back doors we know about, and asking users to change
their passwords."
   I explained that I had convinced Katz at dinner the night
before that to take these steps now would be a disaster for us, and
would probably end any chance we had of catching the intruder.
   "Tsutornu," Claudia snapped, "you've been here a week now,
and I don't see anything in the way of progress."
   "Excuse me," I shot back. "Get this straight-I've been here
about twenty-four hours and it is obvious that you guys were just
thrashing before I got here. I'm busy and I don't have time to deal
with you right now," I said, and abruptly turned back to the con-
versation I'd been having with Andrew.
   Julia thankfully was more diplomatic, and took Claudia aside
to explain the real progress we'd made so far, and what our plan
of action was for the next couple of days. She also learned that
Claudia was upset in part because Pei had her employees on-site
                        "YOU LAMERS!"                         177
around the clock looking at the snoop data, and it was costing the
Well a fair amount of money.
   A while later Julia came back and said it didn't look as if
Claudia was going to recommend that the board throw us out
immediately. The crisis had been averted for the moment, but it
was becoming increasingly obvious that we were going to have to
move forward as quickly as possible or the investigation would be
   Most of our time on Wednesday was spent waiting for a pro-
gram I'd written called Crunch to run. It was designed to take the
filtered packet data we'd accumulated from the night before, sort
through it, and organize it into distinct sessions, which should
make it possible to reconstruct exactly what the attacker was up
to. But Crunch was running slowly, taking twice as long to sort
data as it had taken to collect it. We had managed to speed it up
a bit by playing around with some things in the Well's network
that were broken, but it was a larger filtering net than I'd ever
cast-and it was on a busier computer than I'd ever dealt with
   While I waited, I sat down at Pei's computer console and began
my own hunt through the data we'd collected. Among the hundreds
of stolen files hidden in the purloined accounts we'd found so far
was the credit card database file from Netcom. Among the names
were those of a few people I knew, including a friend of mine who
was a housemate of Castor Fu. Castor was out when I called, but I
left him a messageasking him to read the credit card number to my
friend. It would probably leave him with a queasy feeling.
   I next phoned Mark Lottor, and together we tried to figure out
where his stolen code we had found at the Well had come from,
and when it had been taken. As I described the file he realized it
was a very old version of his Oki code, which meant that it had
probably come from my computer, as Mark had the latest copies.
I went back to poking around the stolen software when Andrew
walked up and saw what I was doing.
   "Hey, wait. How come when I go look at this data, it's botany,
but if you look through it it's okay-it's science?"
   I just grinned.
   As I continued to look back through the earlier data, I noticed
that on that morning we had captured packet data showing that
178                         PURSUIT
a break-in from the Well to Internex had been dropped in mid-
word. In a session from Netcom that began at 7:29 A.M., the
intruder had apparently begun to type the command "uude-
code," but the connection was dropped at 7:31 after he had typed
only uudeco, the first six letters. Minutes later he came back and
picked up exactly where he left off, using the command to decode
and then run a program called 1.Z that made him root on
Internex. But the broken session suggested that the FBI may have
been right in their belief he was running over an unreliable cellu-
lar telephone connection. In either case, we had a precious clue,
a signpost that would appear simultaneously in each of the con-
nection logs of the computer network operators and the tele-
phone companies' phone call records stretching all the way to our
intruder's actual physical location.
   I had been hoping that Julia and I would be able to go out for
a skate sometime during the afternoon, but it was dusk by the
time we took off gliding north along a bike path down
Bridgeway, a road that stretches from Sausalito to Mill Valley. It
was nice to be on skates, but as we started out I felt awkward as
I tried to make the transition from the cross-country skis I had
become used to. It didn't take long, however, to find my rhythm,
and on a long downhill section I looped back and forth waiting
for Julia who doesn't particularly like to go fast downhill. At the
bottom my pager buzzed and although the number was one I
didn't recognize, I punched it into my cell phone and returned
the call anyway. It was David Bank, the SanJose Mercury reporter.
I was busy, I told him, and I couldn't talk to him. I hung up and
thought to myself: Now that I know his number, I'll know how to
ignore hispages.
  We skated out for thirty minutes or so and then turned around.
By now darkness had fallen so we stopped and phoned Andrew,
and asked him to come pick us up. We skated in circles, until he
arrived and drove us to the Samurai, a Japanese restaurant in
Sausalito. Over dinner the three of us talked about where to go
next. It was clear that we needed to move our base of operations,
but I was still skeptical about the value of going to Netcorn or
Intermetrics, and it occurred to me that if the FBI thought our
intruder was actually in Colorado then maybe we should head
that way next.
                         "YOU LAMERS!"                         I 79
  Julia argued against this plan, however, because she wasn't con-
vinced the FBI had adequate evidence to support their belief
that the intruder was actually in Colorado. I pointed out that
since we were seeing the most activity coming from there, it was
worth a visit; if he wasn't there, however, we could figure it out
quickly and move on. Andrew was worried that the Colorado
SuperNet (CSN) systems administrators seemed a bit slow to
catch on, and reminded us of the incident we had learned about
that morning in which the CSN staff had managed to let them-
selves be counter-detected. I decided that I'd call them and see if
they were willing to cooperate with us. We left the restaurant,
realizing there was still a tremendous amount of work to do,
including setting up our second monitoring station at the Well
to give us a backup.
  After we returned to the Well, Andrew called CSN. He talked
for a while to one of their people who was working with the FBI
and then passed the phone to me. I wanted to know whether the
intruder was using the local Colorado dial-in telephone lines or
whether he was coming over the Internet.
   "We've been watching attacks on the Well coming from your
computers and I thought we could find some way to share infor-
mation," I explained to the CSN systems manager.
   "We're working very closely with the FBI," he replied. "Thanks
for your offer, but we have this under control. I've been instruct-
ed not to give you any information, but to ask you to contact the
Los Angeles office of the FBI, and they will pass relevant infor-
mation on to you in a timely fashion."
   "Timely?" I couldn't believe what I was hearing. "But you were
counter-detected this morning!"
   "I know we made a mistake," he replied brusquely, "and we'll
make sure it doesn't happen again."
   I asked him whether CSN now had trap-and-trace in place and
whether they were in touch with their local cellular telephone
company. He said, yes, they had that covered.
   He didn't seem very sincere, though, so as a benchmark test I
asked, "Are you asking the cellular telephone company to watch
for all data calls to see whether he's in the area?" This was some-
thing that was clearly impossible, because there was no way that
a cellular phone company could conduct surveillance on all of its
I 80                       PURSUIT
phone calls. "Oh, yes," he said blankly. Talking to these people
obviously was a complete waste of time, so I hung up. If we went
to Colorado, we were going to have to start from the ground up.
It didn't look like a viable option.
   After the phone call I turned my attention to figuring out why
we couldn't get the new disk working for the second monitoring
station we were trying to set up. Most workstations and an
increasing number of personal computers use a standard hard-
ware connection known as Small Computer Standard
Interface-the acronym is pronounced "scuzzy"-to connect
things like hard disks and CD-ROM drives. Our second RDI
refused to recognize the disk that Pei had lent us, and though
Andrew had tried another cable to connect it, we still had no
luck. Now I took a stab at the problem. Ordinarily, a SCSI bus
needs to be properly terminated-a damping function that
makes sure the signals on the cable don't reflect and interfere
with each other, but after playing around with a variety of dif-
ferent things we realized that when I left the external termina-
tion off the drive it suddenly started working. Odd, but hard-
ware can be like that.
   With all.of our monitoring stations running I returned to the
filter data. Around eight that evening our intruder had been
prowling the Well, following his normal routine of making him-
self root and then hiding his presence with a cloaking program.
He checked briefly to see if Jon Littman had received any new
mail, found none, and so turned his attention to Markoff. As he
opened the mail file he used a standard Unix text search com-

# grep -i itni rnbox

  wait, I thought to myself, this is something we haven't seen
before. He was looking for the four letter string "itni" in
Markoff's electronic mail file. The intruder was trying to be dis-
creet, but to me, it was a dead giveaway: it looked as if Kevin
Mitnick was on the run; he had apparently taken a keen inter-
est in knowing what people might be telling Markoff about
                         "YOU LAMERS!"                         I 8I
him. In this case he was out of luck, for there was no Mitnick
material to be found.

While Andrew and I had been methodically tracking our intrud-
er during the week through the Net, we were now gening indi-
cations from Mark Seiden at Internex that a similar break-in pat-
tern was beginning to emerge there as well.
   I knew Seiden a little because we'd spent some time together at
the annual Hackers' Conferences at Lake Tahoe and other com-
puter conferences over the years. He was also a friend of
Markoff's and Lotter's. With curly black hair, a graying beard,
and wire-rim glasses, Seiden tended to affect what some people
see as the same anti-fashion statement typical of Andrew and me.
You can usually find him dressed in a T-shirt emblazoned with
some technology theme, shorts, a fannypack, and sandals, and he
is seldom without his pager, cell phone, and RadioMail terminal.
A graduate of Bronx Science High School, where he was in the
same class as Bruce Koball, and a one-time researcher at IBM's
Thomas Watson Research Center in Yorktown Heights, New
York, he is another member of the first generation to have grown
up with computers. A skilled Unix hacker, Seiden has had a range
of consulting jobs with some of the nation's biggest online com-
panies. He also has had a good business installing firewalls for all
kinds of companies ranging from Internet providers and software
companies to prestigious New York law firms.
   Seiden took a special interest in the Internex break-in because
his consulting group, MSB Associates, was located in the same
building as Internex in downtown Menlo Park, and MSB also
had its Internet connection provided by Internex, When we first
talked by phone when I returned his call on Tuesday, I sketched
the scenario out briefly, telling him we had seen a large file being
transferred to gaia.internex.net and asking for his help. I also
explained that we were doing everything possible to avoid tipping
off the intruder, that we had growing evidence that our interlop-
er was Kevin Mitnick, and I wanted the file which contained my
personal data removed from Internex quickly, because I didn't
want it propagated all over the Internet. Mark agreed to do his
own surveillance and later on Tuesday talked to Andrew to coor-
I 82                      PURSUIT
dinate the details. After Andrew described the agreement he had
made with the Well not to copy materials, Mark decided he didn't
want to limit his own freedom according to those terms, and said
he would prefer to continue to work independently from us.
   Once he had begun examining the Internex system, he soon
found that an account named brian had been commandeered on
the company's computers which were located on the second
floor of a downtown office building above a barbershop. The
account actually belonged to Brian Behlendorf, a former
Internex consultant who was now working at Wired, doing
Internet and World Wide Web development. When Mark
peeked to see what was stored in the brian directory, he found a
copy of tsu. tgz, the same bundled and compressed file of my
directory that we had found at the Well. Working from his own
computer down the hallway, which was connected to the larger
Internex network by a local Ethernet network, Mark set up his
own sniffer programs to monitor all external connections to
Internex. Since his computer was not obviously a part of the
Internex network, and was being closely monitored, he had told
me he was fairly certain the intruder hadn't broken in to his
machine. He was confident he could use it as an observation
post from where it was unlikely that the invader could know that
someone was dogging his every step.
   As Mark began exploring the Internex computers for Trojan
Horse and back door programs left behind by the intruder, it
took only a few minutes to spot an innocuous-looking program
running on Gaia, their mail-handling computer, called in.pmd.
Pmd would normally be the name of a program known as the
Portmaster daemon, a tiny piece of software that communicates
with hardware devices that would normally connect users dial-
ing in from the outside world to the computer. In this case,
however, it was immediately noticeable because Internex was
not running any Portmasters. The intruder hadn't bothered to
check to see if his ruse made any sense in the context, or maybe
he didn't care.
   Mark took the tiny program apart and discovered it was mini-
 mally camouflaged. Its operation was simple: If someone con-
 nected to port 5553 on the Internex computer and typed "wank,"
 they would automatically become root, with all the power that
                         "YOU LAMERS!"                         I 83
entails. Interestingly, in.pmd existed only in the computer's
memory; there was no corresponding version of the program to
be found on the hard disk. This meant that the cracker had
copied it over to Gaia's hard disk, started it running in the com-
puter's memory, and then erased it from the disk, which made its
presence harder to detect. The program had been left running for
the intruder to use whenever it was needed with the assumption
that no one would notice it.
   Andrew had tipped Mark off to some of the cracker's tricks,
and as Mark continued to investigate he found that someone had
tampered with a standard, but now rarely used Unix system pro-
gram called newgrp, a utility program that assigns a user to a par-
ticular group for organizational or access purposes. The intruder
had replaced the original newgrp with another program that has
the same name, but which secretly had other functions as well.
We were familiar with it, as it is a fairly common Trojan Horse
program that floats around in the computer underground. The
Trojaned version of newgrp allowed the intruder to make himself
root or to pose as any other user on the system. The more Mark
investigated, the more he realized Internex had been thoroughly
penetrated. He discovered a handful of other Trojaned programs
and innocuous accounts with names like "sue," set up and left
lying unused, apparently as a backup in case the intruder found
himself shut out.
   Shortly before midnight on Tuesday Mark rebooted Internex's
computer to flush out any hidden backdoors or secret daemons
he hadn't been able to find, and also erased the Trojaned version
of newgrp.
   By Wednesday, just after 7 A.M., the intruder was back, this time
connecting from escape.com, trying to use his now missing back
door. Failing to get in, he logged-in seconds later to the brian
account. He had changed the password to fucknmc, which had
clearly become a mantra to him. Once inside he checked to see
who was currently logged-in and who had been on the system
recently. He then fetched a copy of the daemon program Mark
had deleted the day before and installed it in Internex computer's
memory, once again deleting it from the disk after he was finished.
   Thirty minutes later he was back from the Well, laboriously
 reinstalling and hiding his Trojan newgrp program that Mark had
I 84                        PURSUIT
erased the night before. As Mark monitored from his computer,
he watched the intruder check through all of the Internex mail
aliases for "mark," presumably to find out where Markoff's mail
was being sent. Not only did Markoff show up, but so did the
name Mark Seiden-but the intruder didn't seem interested.
Shortly afterward the invader checked to see if Markoff's New
YOrk Times Internet address was connected to the Net. He might
have been interested in breaking in to that computer but it did-
n't answer, so he instead altered Markoff's mail alias so that a
copy of all of his inbound electronic mail would automatically be
sent to a mysterious account at Denver University. The attempt
to redirect the reporter's mail failed, however, because it was set
up incorrectly. Almost twelve hours later the cracker was back
making himself root and browsing through all the subject lines in
Markoff's mailbox. Although there was a lot of junk mail, one
subject line did say "Intel stuff," but the intruder didn't seem to
be interested in that topic.                                 .
   Since the illicit programs had been immediately reinstalled,
that evening Mark decided not to erase them again, but instead
to write his own small program that would not only send him
an alert each time someone connected through the covert back
door, but also included a surveillance countermeasure. Since he
received information on where the intruder was coming from
each time, he wrote the program so that it could check who was
currently logged in at the offending site, by running finger
against the attacking computer. As he observed the intruder's
comings and goings over the following days he would see that
although in some cases the intruder would connect to Internex
from the Well, he would most frequently enter via escape.com,
which he learned was a New York City Internet service provider
run by an enterprising high school kid. The listings of current
users that came back frequently included names of people who
were currently connected to the system including Phiber Optic,
and Emmanuel Goldstein. I wouldn't call it a slum, but it was
probably one of the Internet's more seedy neighborhoods.

Everywhere we looked there were more signs pointing to Kevin
Mitnick, but my immediate challenge was making sure that the
                           "YOU LAMERS!"                            I 85
  filters provided clues about his location, and on Wednesday after
  midnight at the Well as I ran through the elements of our moni-
  toring setup in my head, it occurred to me to ask Andrew
  whether he had time synchronization running. Time synch is a
  computer network utility that ensures that each computer's clock
  matches the others in the network. It is a useful tool for all kinds
  of computer-related activities, and it's essential for computer
  security work. In big computer networks you may have hundreds
  of people logging on and off every minute and thousands of
  events taking place. The only way to ensure accurate reconstruc-
  tion of this activity is to be absolutely certain that the clocks
  match across the network.
     "I assume it's in place," was Andrew's response.
     Assume? We checked and of course the Well wasn't running
  time synch, neither were we.
     What that slip meant was that all the data we'd collected from
  the night before would be difficult to use, at least for traffic analy-
  sis. If the clocks weren't synchronized, it would be far more diffi-
  cult to match events taking place on different machines, a neces-
  sary step to tracing someone who is connected through a string
  of computers on the Internet.
     "I don't ever want to hear the A-word again," I told Andrew.
     He seemed miffed. It had been a long, rough week of twenty-
  hour days for him, and he was bearing the brunt of everything
  that was going wrong. But it was wrong to assume that because
  we ran time synch back at SDSC that everyone else did. From my
  point of view running time synch is an absolute, non-negotiable
  requirement, for the essence of everything I do is related to time.
     I spent some more time that night working on tools I could use
  to look at the data we were collecting. Earlier in the evening John
  Gilmore had called my cell phone looking for Julia, and she had
  disappeared into another room to talk to him and was gone for
, several hours. She had been working on a tool for searching
  through the data that we needed as soon as possible, and when it
  didn't get done I picked up the project and finished it. When she
  came back she was distraught, and the two of us went outside and
  took a walk along the Sausalito waterfront.
     "John wants me to go to Wylbur Hot Springs this weekend,"
  she said. It was to be the weekend together that Julia had agreed
I 86                        PURSUIT
to before she had come to the Vanguard conference in Palm
Springs. Wylbur Hot Springs was a funky, sixties-style retreat
north of San Francisco. We talked about it as we walked along the
waterfront near the pilings and the houseboats. In my mind it
seemed that there would be much more to the weekend than
what appeared at face value.
   "It's a place we used to visit, when things were better between
us," Julia said as we walked a little farther.
   Even though the idea was supposed to be to say goodbye, we
both realized John had something else in mind, and Julia found
this disturbing.
   We strolled on in silence. I didn't have an answer.
   "We have to go back to work and finish up," I finally said.
   It was after 3 A.M. when we drove back to Dan Farmer's house.
I felt a growing urgency to do something to find the intruder
quickly, but I still wasn't sure which direction to move in. One
thing I could see clearly, though, was that we didn't gain anything
by remaining at the Well.
   Our room at the back of Dan's house contained a bed, a Sun
workstation, and many shelves full of science fiction books. It
also had a water fountain with an artificial stone basin, several
tiers of dark stonelike material with pebbles strewn about. I was
exhausted, and I barely heard the water cascading over the sides,
making a soft burbling sound, a white noise masking the whin-
ing of the workstation sitting in the opposite corner of the room,
before I fell into a deep sleep.
I awoke on Thursday to the sound of falling water and although
the bedroom was still dark, I saw that it was late morning.
Several of the household cats were careening about, and in the
dim light I could make out Dan's housemate's collection of sin-
gle malt scotches dotted around the room. I knew the time had
come to choose the direction for our next move.
   Andrew had been at the Well for a week, and now we had
some real data, and a few potentially solid leads. But our intrud-
er was still running amuck, and we needed to get moving. With
Colorado SuperNet an increasingly unfeasible option, I had to
face the idea of doing our observing from Netcom. As one of the
nation's largest InterNet service providers, it was going to make
our search like looking for someone in Grand Central Terminal.
Being blocked at CSN was frustrating to me and I spent several
fruitless hours in the early afternoon trying to see if there was a
way to circumvent the L.A. FBI's roadblock in Colorado.
   Afterward, because Julia was feeling agitated, we talked for a
while. It was clear to both of us that John was counting on the
pleasant memories of Wylbur Hot Springs to weaken Julia's
resolve to leave him.
   ''I'm afraid I won't be able to maintain my sense of perspective
when I'm with him," she said. Julia was uncertain about her abil-
ity to stay independent from John and she was worried about
being sucked back in to their relationship. "This is going to be
I 88                        PURSUIT
difficult," she added. "I want to make sure I get enough rest
   It was now the middle of the afternoon, and I decided I didn't
want to go back to the Well. For one thing the Wired photograph-
er I was trying to avoid was staked out there, waiting. Andrew had
paged me several times during the afternoon, and I kept putting
him off with ''I'll be there," but it had become clear that we ought
to accept Netcom's offer of support. We'd been seeing traffic from
there for a long time, and a vantage point at their headquarters
would give us a nationwide listening post on their nationwide net-
work, and possibly an "upstream" location that would place us clos-
er to our attacker. Besides, we had several telltale events-the
abruptly interrupted sessions-that we might be able to use to pin
down the intruder's identity on the Netcom system.
   I contacted Rick Francis, the Netcom VP of software develop-
ment who had been involved in the telephone conference call at
the Well on Tuesday, told him my plan, and asked if his offer still
held. I apologized for calling him at the end of the business day,
but he didn't seemed to mind, and he told me that his staffwould
still be around for a while to talk to us.
   Before leaving I called Andrew and had him read me the pre-
cise times ofseveral events corrected with his best estimates of our
times-synch error, so that we would have something to correlate
with Netcorn's records. It was almost four in the afternoon when
Julia and I picked up burritos at Zona Rosa on Haight Street and
drove down 1-280 to San Jose in her Mazda. I figured one way or
another we'd be able to get my computer from the Well later on.
   Tracking a stretch of the San Andreas fault, 280 is considered
by many to be the world's most beautiful freeway. It's a nice
description, although I've always found it to be an oxymoron.
Tucked up against the Santa Cruz mountains, 280 runs down the
center of the Peninsula, and is actually Silicon Valley's
Mulholland Drive. On the way south the highway winds through
Woodside, Portola Valley, and Los Altos Hills, where the Valley's
new and old money commingle. As you drive by the thousands
of acres that were once Leland Stanford's farm, you can still see
dairy cows grazing not far from 3000 Sand. Hill Road, the brain
center of the Valley's venture capitalist community, which is the
principal beneficiaries of what has been described as the largest
legal accumulation of wealth in history.
                             NETCOM                            I 89
  As we drove I ate my burrito and phoned Kent Walker to let
him know about our next step. I told him about being stymied
in Colorado and asked about the limits of the Electronic
Communications Privacy Act, which made it illegal to intercept
cellular telephone calls. Even if it was illegal to listen to voice
calls, I wondered, would it be a violation of the law simply to
look for the presence or absence of a data carrier being sent by
modem over a cellular telephone call? He replied that as long as
we weren't deciphering the content of the data, such an intercep-
tion would probably be legal. At this point my question was only
hypothetical, but at some point it had occurred to me that that
course of action might be our only option.
   Twenty-five kilometers farther south in Cupertino, 1-280 pass-
es Apple Computer's new research and development campus.
Here in 1993 former Apple chairman John Sculley planned to
install himself as the company's chief technical officer only to be
deposed in a boardroom coup much like the one in which he
ousted Apple's original visionary, Steve Jobs, eight years earlier.
From Cupertino the highway arcs through the heart of Silicon
Valley, offering endless vistas of low-slung semiconductor fabri-
cation and computer design and assembly plants.
   Netcom itself is located in a twelve-story steel and glass tower
across the street from the Winchester Mystery House just off 280
in San Jose. Now a tourist attraction, the Winchester home was
designed by the paranoid widow of the inventor of the
Winchester repeating rifle and is full of hidden rooms and secret
passages that lead nowhere. The Winchester name was later bor-
rowed by IBM's disk drive manufacturing division, located at the
very south end of the Valley, for the first modern hard drive.
   After we stepped through the front doors and began our search
for the Netcom offices, it felt as if we had made a wrong turn and
had found ourselves instead across the street at the Mystery
House. We ended up going down one flight of stairs, reversing
directions after sticking our heads into a lobby, and then heading
up several flights before we finally found Rick Francis's office.
   Sociologically speaking, Silicon Valley can be divided into
"techies" and "suits." The difference between the two is often
only that the suits know how to dress and have managed to work
their way out of the engineering ranks and into management.
Francis was clearly a suit, dapper in a buttondown shirt, tasseled
190                         PURSUIT
loafers, and a patterned sweater, the typical uniform of product
marketing managers and engineering vice presidents throughout
the Valley. Dealing with an outsider on a computer security issue
was obviously new territory for him, and though he wanted to be
helpful he wasn't quite sure what to expect from me and so was a
bit guarded.
  After I quickly filled Francis in on what we knew from our time
at the Well, we went upstairs and rounded up two members of
his technical staff One of them, John Hoffman, a systems
administrator, was a quiet engineering type who configured and
maintained Netcorn's computer systems. The other, Robert
Hood, was a network administrator who appeared to be a gen-
uine hacker who really knew his stuff. He was calm, knowledge-
able, and didn't seem arrogant about his abilities. He was also the
stylistic counterpoint to Francis. Plump and clean-shaven, Hood
had masses of curly dark hair that fell twenty centimeters below
his shoulders. He was wearing a faded black Metallica T-shirt
adorned with a grinning skull, blue jeans, running shoes, and had
an alphanumeric pager hanging from his belt. I liked him imme-
diately. Robert was the classic Silicon Valley hacker ~ho gen-
uinely liked his work. He'd grown up with Netcom from its first
days as a local Internet provider.
  After Francis told Hood and Hoffman, "Give them any of
your time and any equipment they need," we found a confer-
ence room and got to work. I made it clear that our goal was to
locate our quarry as quickly as possible and keep moving
upstream until we pinned him down. Once again, I reviewed
the data from the Well, and pointed out that the connections
we were interested in were coming repeatedly from Netcom and
CSN. I also explained that we had a growing suspicion that we
were dealing with Kevin Mitnick. The Netcom crew already
knew who he was; it seemed that he had caused a lot of trouble
for them in the past.
   I showed Robert the list of events that Andrew had read to me
and asked him if it provided enough information to help figure
out what account the intruder might be using at Netcorn,
   "No problem," he answered.
   During the meeting Robert and I did most of the talking as we
assessed the obstacles in doing packet filtering at Netcom. I asked
                           NETCOM                           I 9I
about their internal network, details of recent break-in incidents
and what kinds of precautions they were taking. I also asked
about the stolen credit card numbers we had found in the dono
account on the Well, and it turned out that these had initially
been stolen almost a year earlier and had been floating around in
the computer underground for some time; their existence had
been mentioned the previous year in 2600 magazine. Francis said
that initially, Netcorn had no firewall protection, and customer
data had been kept on computers that were relatively unprotect-
ed. The oversight had been a costly error, and they knew it. He
wanted to know if we had a copy that was taken after mid-
January of this year. If the credit card data had been stolen again,
it meant they had a huge problem.
    It ended up being a short meeting, which impressed me, as we
had been able to circumvent most of the social niceties and got
down to business immediately. Afterward Francis said that he
wanted me to listen to a tape, and he took us into a room near
his office and played a recording made from a conversation
between one of the system crackers who had been plaguing
Netcom and a technical support person. Francis was obviously
curious to see if the voice sounded similar to one in the messages \
on my voice mail system in San Diego. The taped call involved
the Netcom person talking to the cracker about his motivation
for breaking into their computers and asking about some of his ..-J
methods, but Julia and I agreed that the voice on the other end
of the line sounded nothing like the one on my voicemail.
    It was almost 6:30 P.M., and Francis apologized for having to
leave early. He would have loved to stay around and observe, he
 explained, but he'd scheduled an essential business trip for the
 next morning. Before he left, though, he gave a final authoriza-
 tion to our pursuit.
 cc "Re~ember, anything you need, we'll pick up the.tab," he said.)

  And if you have to travel somewhere to track this guy down,
 Netcom will pay."
    After locking horns with Claudia, and being second-guessed by
 everyone at the Well, Netcorn's unconditional support was a wel-
 come relief. For the first time I began to feel we had a reasonable
 chance of finding our data thief.
    Julia and I and the two Netcom systems guys went upstairs and
I 92                        PURSUIT
squeezed into Robert's tiny office, which barely had room for a
Sun workstation and was crammed full of technical manuals.
   The list I had taken from Andrew over the phone gave us pre-
cise start and end times of sessions coming from Netcom to the
illicit accounts on the Well, so our challenge was to find out if
there was a single user who had been logged on to Netcorn at all
of the times on my list. We had one critical clue: the time of the
dropped connection on the Well should correspond to a similar
log-out on Netcom, an event that should stand out in the moun-
tains of accounting data. Moreover, finding that our attacker
turned out to be a single person, rather than several people or a
gang sharing one account, would greatly simplify our task-there
would be only one location to hunt for. I was counting on
Occam's Razor, the principle in science and philosophy that
when competing theories exist to explain some unknown phe-
nomenon, the simplest explanation is preferable.
   Robert took a seat at his workstation and the three of us crowd-
ed around him and watched as he searched through his files for log-
ins and log-outs that had occurred at particular times. I could
immediately tell that he qualified as a genuine Unix wizard. He
never hesitated at the keyboard, and commands just flowed from
his fingers. When I asked questions he wouldn't pause to remem-
ber how to find a particular piece of information, but the results
would just appear almost instantly. Robert was also committed to
catching our intruder. "This guy has been really plaguing us," he
said. "I've begun to take this personally. If you find him, I'll be
there with you. Rick Francis and Bob Rieger, our chairman, will be
there, too. They're totally pissed off about this."
   He was clearly jazzed by our arrival. With Netcorn's continuing
expansion in different cities his plate had already been quite full.
Now he was looking forward to an adventure in which he wouldn't
have to balance his official system administration work against the
hunt for the intruder.
   To look for a match between the Netcom and the Well data, he
needed to hunt for information on the 23 Sun SPARCstations
that made up Netcom's online service. Robert had a script that
would search through system accounting records on all the
machines going back to January 1, but it would take a while.
   While his script ran, Robert started to tell me about Netcom's
                              NETCOM                             I 93
internal network. He explained that the 23 SPARCstations were
all connected to a Fiber Distributed Data Interface, or FDDI,
local network ring. Also connected to this ring were the routing
computers that provided connectivity to the Internet, as well as
their own transcontinental T-3 network, capable of moving
almost 45 million bits of information per second. This backbone,
in turn, was connected to a nationwide web of T-1 data lines,
linking both their high-speed data customers and their local
Points of Presence, or POPs, back to their San Jose network hub.
   Instead of having a single 800 or long distance phone number,
most national Internet service providers place POPs with small
pools of dial-up modems in dozens or even hundreds of cities
around the country. It was this ability to establish a private data
network-one that bypassed the standard public telephone long
distance network-that created the economies of scale that made
it possible for Netcom to do business as a nationwide Internet
service provider with local dial-ups in even fairly small cities
around the country.
   Perhaps Netcorn's efforts to make their network easy to access
would work in our favor. While we had never seen the intruder
using the Well's telephone dial-up lines to come directly to the
Sausalito online system-he had always connected over the
Internet-Netcom had local dial-up lines in 51 cities around the
country-if the intruder was sloppy, it was possible he would tip
his hand as to his actual location with a phone call to a local
Netcom number. Trap-and-trace information from the phone
company might then allow us to pin him down even if he was
using a cellular phone.
   We talked about what would be involved in setting up moni-
toring on a network of computers that was larger than any I'd
ever faced before. What I needed was a single point from which
we could get access to all of the packets that were percolating
through Netcom. Dealing with the Well had been like standing
at a street corner on Main Street in a small midwestern town and
intercepting all the red Fords or all the cars with California
license plates as they went by and photographing the driver of
each one. Netcorn, in contrast, would be like coming to Los
Angeles and doing the same thing on the Santa Monica freeway.
   It turned out that there was a single choke point in their network.
I 94                        PURSUIT
That was the good news. The bad news was that it was on their
main FDDI ring. FDDI is a very high-speed computer network
standard which transmits data at 100 million bits per second, ten
times faster than the Ethernet network we'd been dealing with at
the Well. Monitoring this network would require additional
hardware and special software, as the Ethernet monitoring tools
we'd employed at the Well were useless here.
   By now the data had been collected on user log-ins and Robert
began plowing through them, looking for a match. After a while
it became increasingly obvious that there was a single account
that matched with the log-ins of the trespasser at the Well in each
   A user of an account named gkremen appeared to be our culprit.
There were severallocal log-ins this month from San Francisco, but
every remote direct dial-up access to gkremen came exclusively
through their remote POP in Raleigh-Durham, North Carolina.
   "I'm sure it's him," Robert said, though I was hesitant about
jumping to a conclusion prematurely, particularly because we had
only four data points, three log-ins to the Well and one ftp ses-
sion, to work from. We looked more closely at gkremen. Who
was this guy? We found Netcom account information that indi-
cated that gkremen was a legitimate user, not a made-up account
like many of those we'd found on the Well and Internex.
Gkremen was leasing a high-speed network connection from
Netcom directly from his computer's location, but he also had a
secondary account on Netcorn's systems, known as a "shell"
account. It looked like the real gkremen used the account on rare
occasions, and as we examined the connection records it became
increasingly apparent that his account had been taken over.
   Robert scanned through gkremen's home directory, and it was
pretty boring except for one thing that caught his attention: a
small program called test 1. He explained that it was a version of
the telnet program which didn't log its usage. Normally when
anyone uses the standard Netcom telnet program to connect to
another computer, the user's name and remote computer name
are recorded. Robert had already begun working on a modifica-
tion for Netcorn's operating system so that the record-keeping
function couldn't be circumvented. Obviously, someone had
commandeered gkremen's account and was using it secretly.
                             NETCOM                            I 95
   It was looking more and more as if we'd struck gold. Scanning
through the log-in records for gkremen, we could also see con-
nections from familiar sites like escape.com and csn.org. Raleigh
seemed to be his favorite, though; in the past five days he'd come
through there 26 times. He'd been on almost every day, includ-
ing a few sessions that morning.
   I thought I remembered some of my friends who live in
Raleigh complaining about the quality of their phone service.
   "Robert, do you know who the phone company is near
Raleigh?" I asked.
   "'T ah , " h e rep 1· d "I's GTE"
     re                ieo, t      .
   Groan. "Oh no, I was afraid of that."
   GTE had a reputation for having lax security. Their central
office switches were notorious for being commandeered by
phone phreaks who would secretly reprogram them to get free
phone calls and often play esoteric and nasty tricks. Our task
would be made much more difficult if our intruder had also
managed to tamper with the telephone company's equipment,
but that was a hurdle we wouldn't have to face for a while.
   Potentially the Raleigh discovery was a significant break. If the
intruder was simply connecting to Netcom from the Raleigh pap,
our monitoring operations might be greatly simplified. At each of
their POPs Netcom used an Ethernet network to connect equip-
ment ranging from Portmasters to routers. If we could find a sin-
gle local site on the periphery of Netcom's national data network
we would avoid having to build an FCCI-monitoring system, and
sort through the vast quantities of data that ran along the FDDI
network backbone here in San Jose. We began checking airline
schedules to see how quickly we could get someone to Raleigh, and
at the same time I called Kent to ask him to get a trap-and-trace
order for the Raleigh POP.
   "I can't do it tonight, because it's too late," he answered. "But
I'll get it set up first thing in the morning. Who is the phone
company?" I told him, but he didn't seem to have the same reac-
tion to GTE that I did.
   While I was talking to Kent, Robert wrote a simple script so
that every time the gkremen account was used, an alert would be
sent to his pager telling him which Netcom POP the call was
coming from.
I 96                        PURSUIT
   It was almost 8:30 P.M., and Robert's pager went off almost
immediately after he'd finished installing the alert, but this time
it was with bad news. Gkremen had logged in, but he was not
coming into the Netcom system through their Raleigh POP, this
time he was coming through Denver!
   Damn, I thought to myself, he's been incredibly consistentfor the
past five days and now we show up and he shifts his location. It
meant that we couldn't be sure he would go through Raleigh, and
we would therefore have to look at all the data from across the
country in the Netcom network to trace him. I wondered briefly
if he'd been spooked or was actually somewhere else. Although it
was possible that he was dialing in to different POPs in an
attempt to be covert and hide his reallocation, Robert mentioned
that they were having technical problems in Raleigh and it was
also possible that the intruder was just phoning to a different
POP to get a working modem line.
   As we watched, Robert used a diagnostic utility at the POP to
peek at gkremen's keyboard session. Although the software was-
n't meant for monitoring a live session, it worked in that capac-
ity-sort of. As the person who was using gkremen's account
typed, Robert would click on his mouse, and the contents of a
small memory buffer from a Portmaster at the Denver site would
be displayed on his screen showing us what the intruder was typ-
ing. Unfortunately the buffer could only display sixty character
snippets of the activity going in each direction, which meant
that we could see most of what the intruder was typing at his
keyboard, but only an occasional glimpse of what he was actual-
ly seeing on his screen. We also encountered another problem
that made it even harder to see what was happening clearly. At
the time Netcorn was grappling with a software bug in its largest
Cisco routers. These are the computers that are responsible for
directing the billions of packets of data per day that were circu-
lating around the FDDI network ring and sending them to the
right destinations in the Internet. Every thirty seconds or so, the
entire network would have a miniseizure, meaning that we lost
more keystrokes. It was a little like wiretapping someone while
playing the 1812 Overture loudly in the background.
   Despite the dropped packets, however, we were still able to
form a rough idea of what he was doing. We watched as he tried
    to break in to a computer at CSN, apparently without success,
                                                                   I 97

I   and then he turned to another computer at the Colorado facility
    and tried to edit one of its system configuration files, but found
    that it was a read-only file, and could therefore not be tampered
       We watched as the trespasser next used the file transfer com-
    mand to connect to the public archive computer of CERT, the
    government security information center.
       I started laughing. "Looks like I was right, the anklebiters are
    reading technical manuals," I said.
       He was searching through CERT's files for the word "moni-
    tor," and his intentions were obvious: he was trying to figure out
    how to insert a small network-monitoring program back inside
    the operating system of one of the CSN computers. The pro-
    gram, known as NIT, or Network Interface Tap, is a standard part
    of the computer's basic operating software, but it is usually
    removed for security reasons. If he could reinstall it in the oper-
    ating system he would be able to secretly capture passwords and
    other useful information. He found what he was looking for in a
    file named 94:01.ongoing.network.monitoring.attacks. The file
    gave instructions on how to turn off the monitoring software,
    and now he was trying to figure out how to turn it back on.
    Ironically, the CERT file wasn't even a recent advisory, but was
    actually more than a year old. But he nevertheless attended to it
    as diligently as someone following a recipe in a cookbook, work-
    ing right under the noses of the CSN system administrators.
       After seeing that our intruder was now coming from Denver, I
    placed another call to Kent to tell him we'd now need trap-and-
    trace there as well as in Raleigh. While Robert and Julia stayed
    entranced with the snippets of the intruder's sessions, I began
    thinking about how we were going to put a monitoring operation
    together to allow us to actually trace him. Each of Netcom's POPs
    had banks of modems plugged into devices called Portmaster
    communications servers, manufactured by Livingston Enterprises,
    a Pleasanton, California company. The Portmasters let dial-up
    users access Netcom's computers over its own network. Our
    problem was that the Portmasters, unlike other models, com-
    bined the separate sessions to each computer into a single stream
    of data and we were unable to break them out individually.
I 98                       PURSUIT
Robert knew Livingston's founder, and said he would make an
emergency call to ask him if he could help us in unravelling the
   Figuring out how to monitor the FDDI ring was our next
problem. This would require a fast computer, interface card and
a concentrator to hook the machine onto Netcorn's ring.
Unfortunately Netcorn didn't have any spares of this hardware.
Assuming we could find the hardware, we would still need source
code to the software driver for the card so that we could modify
it to let us monitor the ring. I remembered I had FDDI software
on a backup tape in my house in San Diego, but since nobody
there had a spare key, it wasn't going to do us much good.
   I paced around Robert's crowded office trying to think about
where we could get a FDDI concentrator to hook a monitoring
computer onto the Netcom ring.
   I was racking my brain trying to think of someplace where I
might get the necessary equipment late at night in Silicon Valley.
I couldn't just walk in the door somewhere and help myself, and
they wouldn't have what we needed at Fry's, the Valley's techie
supply store famous for selling everything from computers to
potato chips-FDDI concentrators typically cost many thou-
sands of dollars.
   I suddenly realized I knew just the person.
   1 called my friend Soeren Christensen, an ATM network wiz-
ard at Sun and someone whom 1 had worked with. He was still
at the office when 1 phoned and, after explaining our predica-
ment, 1 told him it was vital we have a working monitoring sta-
tion in place by 7 A.M. the following morning, the time our
intruder usually reappeared each day.
   "Soeren, do you remember that FDDl concentrator that used
to be in the ceiling of your lab in Mountain View before you
moved to Menlo Park?" 1 asked him. "You didn't hang on to it
and have it lying around somewhere now, do you?"
   "1 think I can find what you need, Tsutomu. I think 1 remem-
ber where it is," he answered. "1 can probably dig up some extra
hardware too. I'll look around."
   "That would be great," 1 said. "Where can we meet you?"
   It turned out Soeren was planning to have dinner with his wife
at a microbrewery in Sunnyvale called the Fault Line, not far

                                    NETCOM                            I 99
       from Netcom's offices. "We'll get organized here and meet you
       there in a little while," I told Soeren.
         When I got off the phone Robert and Julia were still watching
       the attacker's antics and it took me a while to pull them away to
       make sure we got to the restaurant in time. It was now almost
       9:40 P.M., and the brewery was scheduled to close in twenty min-
       utes. We decided to take one car to dinner, since we were all plan-
       ning to return and spend the rest of the night setting up our
       monitoring operations. Julia's Mazda was stuffed with my ski
       gear, and so we piled into John Hoffman's shiny blue-green
       Mustang. Both Robert and Hoffman had what looked like
       brand-new American muscle cars. While the standard issue
       Silicon Valley engineer's car is usually a BMW or a Saab, both of
       Netcorn's technical engineers must have had a little of San Jose's
       remaining native culture in them. It felt a little bit like American
        Graffiti, George Lucas's 1973 account of the early sixties in a
       California Central Valley town where life still revolved around
       cars instead of computers.
          The Fault Line is one of dozens of microbreweries that have
       sprung up in the Bay Area during the last decade. An upscale
       replacement for the beer-and-burgers taverns of an earlier era, the
       micro breweries have a more sophisticated California cuisine as
       well as a selection of exotic beers, which are brewed in large vats
       usually set off beyond glass partitions at the back of the building.
          Both Julia and I were intrigued by the beer list, but conceded
       that after a couple of glasses there would have been no way we
       could function through the night, which looked like what was in
       store for us.
          Soeren and his wife Mette had already arrived by the time the
       four of us showed up. I noticed the waitress brought Mette an
       order of horseradish mashed potatoes. Only in California, I
       thought. While I told Soeren what we were up to, we all tried to
       relax, because we knew this might be the last break we would
       have for a while. Over dinner we talked about the monitoring
       system we needed to set up and our problem in getting a com-
       puter fast enough to keep up with the Netcorn's FDDI ring.
       Soeren, who was one of Sun's best designers of networking gear,
       said that he'd been able to find enough random hardware parts
       for us to custom-build a computer. He also thought he had the
20 0                        PURSUIT
FDDI driver source code on a backup tape at his apartment,
which was close to the restaurant. So we arranged for Julia to go
back with him later and pick it up.
   Afier dinner we all stood in the parking lot as Hoffman backed
his car up to the trunk of Soeren's car.
   "This is like a Silicon Valley drug deal," Julia said. Everyone
laughed nervously.
   Of course, in reality it was unlikely anyone was even giving us
a second glance. Half the companies in the Valley probably began
with the salesguys working out of the trunks of their cars. Soeren
handed me two carrying bags full of random gear, including con-
nectors, memory, a processor module, and various interface
cards. I looked at it and said, "Gosh, you shouldn't have gone to
the trouble of tearing it apart, you should have just brought the
whole computer."
   As soon as we drove back to Netcom, Hoffman began assem-
bling the new monitoring computer, placing little green stick-on
dots on all the Sun equipment so we could identity it easily. It
was after 11 P.M. when I paged Andrew. He was over in Berkeley
having dinner with Mark Seiden at the Siam Cuisine, the first,
and some say still the best, of the East Bay's Thai restaurants. We
had agreed to give Mark some of our monitoring functioning to
make it easier for him to track the intruder at Internex.
   ''Andrew, I need you to go back to the Well and get my RDI
computer and bring all of our software monitoring tools down
here to Netcorn," I told him. "It's going to be a long night,
because we need to have monitoring in place by the time he's
active again tomorrow."
   Our interloper would usually begin his day at about 7 Pacific
time and then log-on intermittently throughout the day. He
would usually vanish for a few hours at around three in the after-
noon and then come back with a vengeance and frequently
remain active until well past midnight. It was increasingly clear
that whoever was on the other side of our computer screens was-
n't a casual anklebiter, but an adversary who had a deep obsession
with whatever he was doing.
   Julia returned around midnight with Soeren's FDDI tape,
and it took some time to find a proper tape drive to read it.
When I finally looked at his software my heart sank. It was
FDDI driver software source code all right, but it was written
                              NETCOM                             20I
for Sun's Solaris 2 operating system. Netcorn was runnlllg
Solaris 1. It was hopeless.
   I had hoped to be able to insert Soeren's software easily into our
monitoring computer. If we'd actually had the source code this
would have been fairly straightforward. I had wanted to use my
modified Berkeley Packet Filter software because it was written
efficiently enough to keep up with the torrent of data packets
that were flowing through the Netcorn fiber-optic ring. Now we
were going to have to use another strategy.
   .AB we pursued our various tasks the intruder came back on-line
at 12:40 A.M. He was still logging in from Denver, and he was still
messing around with the CSN computers. A short while later
Robert saw him break into fish.com, Dan Farmer's computer. He
watched as the intruder looked through Dan's mail for any occur-
rence of two different text strings, itni and tsu. The first one
meant he was certainly still looking for the word Mitnick, and the
second was probably for me. If my opponent was indeed Mitnick,
he had now taken an even more consuming interest in me. After
a while he was back on the Netcom computers, this time trying to
find out where Rick Francis's mail was being routed.
   Andrew showed up about two in the morning with our hard-
ware and software, and he immediately set to work trying to fig-
ure out how to install the Berkeley Packet software into the
FDDI driver software from scratch. Without the source code I
was pretty sure it wasn't going to work, but Andrew was opti-
mistic and he set to work seeing if he could make it work by trial
and error.
   Most of the Netcom staff had departed hours earlier, leaving us
alone among the partitioned cubicles in windowless rooms. The
only other people who were around were some telephone
installers at the other end of the floor, putting in a new PBX in
the Netcom machine room. The company looked like a typical
Silicon Valley business in the midst of hypergrowth. As quickly as
these organizations move into new quarters they tend to outgrow
them. Everything seems to be in flux. Unfortunately, another
trademark of the Valley is things tend to collapse as quickly as
they expand.
   By 3 A.M. all of us were already fairly sleep deprived, and
Robert, Hoffman, and Julia kept making runs for the soft drink
machine that was set in an open space across from Robert's office.
202                         PURSUIT
Caffeine has never had much of a positive effect on me. After a
while "Make another selection" lights began showing up next to
the different drink buttons.
   "We're going to run out of everything with caffeine pretty
soon," Andrew said.
   I paced back and forth between Andrew, who was hacking the
FDDI software; Robert, who was monitoring network opera-
tions; and John Hoffman, who was working in the Netcom
machine room still bringing up our new monitoring station.
   Despite the complaints about Netcorn, I was actually quite
impressed by their organization. I walked into their machine
room and saw rows and rows of SPARCstation server computers.
Everything was remarkably neat and professionally set up. It
appeared that the engineering of the system was well thought out.
   It was almost three-thirty in the morning when we finally
turned on the new computer which had been set up behind the
locked door of Netcorn's machine room. Hoffman named it
Looper, a reference to the FDDI network which was set up as a
ring that the packets flowed around.
   Andrew hadn't been able to insert BPF without the source
code, and we were rapidly running out of time for further exper-
   I thought about our other options. We had two different FDDI
cards from Soeren: one built by Sun, and another from a compa-
 ny called Crescendo. I was pretty sure that the Sun card with its
standard driver software, even running on a fast SPARCstation,
wouldn't let us filter packets quickly enough to keep up with
 Netcoms FDDI ring under full load. The Crescendo card and
driver had a better reputation for performance than the Sun card,
 but I didn't know by how much.
   I tried the Crescendo card first. I was hoping that it would
function well enough so that even though we weren't able to get
 the BPF software to work, NIT would do the job. NIT is slow,
 but maybe the speed of the card and SPARCstation 10 could
 make up for the inefficiency. If this didn't solve the problem, the
 only other option was going to be to think of something else
 clever, which I hadn't figured out yet.
    Once it was in place, it took only a couple of minutes to real-
 ize that it wasn't anywhere near fast enough, and that we would
 be fighting a losing battle as the East coast came on line. Around
                             NETCOM                             203
five or six every morning, the number of packets flowing through
their FDDI network would start to soar as people on the East
coast first logged in to check their mail and get their net fix. From
his computer Robert watched the display that monitored the
number of packets coursing over the FDDI network backbone. It
read about 4,000 packets per second.
    "That's about as low as it ever gets," he said.
   Andrew, meanwhile, was watching looper's performance.
   "Tsutornu, this isn't cutting it," he said. Netcom's network was
barely ticking along, and we were already losing one percent of
the packets that were flowing past our monitoring station.
   "This isn't acceptable," I complained to no one in particular.
   We decided to try the Sun FDDI card, but it turned out to be
even slower than the one from Crescendo-and buggier. The
Crescendo card was reinstalled, and we resumed our attempt to
monitor the network.
   It wasn't. As we watched the display on the SPARC station we
saw that it was running at about 70 percent of capacity. Shortly
it got worse. While we were sitting watching, the the number of
data packets being dropped started to spiral upwards as the net-
work load began rising. I had a vision of people all along the East
Coast, still in their bathrobes, holding mugs of coffee, going into
their dens and logging in to Netcorn. I wondered to myself, Do
they have any more ofa lifethan wedo? Well at least they gota good
night's sleep.
  "Tsutomu, in a little while we're going to be up to twenty thou-
sand packets per second," Robert said.
  It seemed obvious that we needed to do something that would
take only a couple of minutes and would work, even if it was a
kludge-computer jargon for a Rube Goldberg device.
  "It might work if we put something in front of NIT," I said to
Andrew. "I could write a quick and dirty prefilter that will sort
packets before they even get into NIT."
  Andrew nodded in agreement, but at this point I'm not sure he
really cared. He was stretched out in an office chair across from
where I was sitting, and seemed already to be half asleep.
  I thought about the problem a little further and read through
the source files for the operating systems I had in order to try to
understand what was happening between the system software
and NIT a little better. It seemed like a tiny prefilter would be
    204                        PURSUIT
    fairly fast-not blazingly fast, certainly, but I hoped it would be
    fast enough to be able to deal with the number of packets flow-
    ing past our monitor machine even at Netcorn's peak loads. My
    program would be the crudest of solutions, a little piece of very
    low-level software that would sit in front of NIT and discard
    most of the packets before they ever got to the cumbersome, inef-
    ficient program. I called it sniL-foo and wrote it without even
    bothering to use an editor. I simply copied each line I wrote into
    a file and then compiled it so that it c~uld be run by the com-
       I sat down in front of the RDI and wrote as quickly as I could
    while Andrew watched over my shoulder. After I finished I
    turned to him and said, "Did I do anything obviously wrong
       He took a cursory look over my code to see if any of it had the
    potential to crash. My program was designed to filter for up to
    eight separate network addresses and reject all the other packets.
    If it worked correctly NIT would have to deal with no more than
    a few percent of the packets that were flowing around the FDDI
       After Andrew inspected the code I compiled it on the RDI
    and it seemed to work. We copied the program onto a floppy
    disk and carried it into the machine room where we put it on
    looper. It was an ugly kludge, but at this point there was noth-
    ing to lose.
       I was exhausted, but still managing to function under the
    pressure of knowing that we might see our opponent at 7 A.M.
    I fumbled around for a while, trying to correctly patch my pro-
    gram into the operating system kernel. It took several attempts
    until I figured out what I was doing wrong. It was almost 6 A.M.
    when packets began trickling into files for us to reconstruct
    later in the day. By now Netcorn's network was coming alive. I
    ran a few test cases and everything appeared to be functioning
    properly and the load on the machine was manageable. Then
    Andrew and I spent some time configuring the filter. With
    everything in hand for the moment, I left the machine room
    and went to see what had become of Julia. She was still plan-
    ning on going away for the weekend with John and about an
    hour earlier she had bailed out and gone and curled up under
                            NETCOM                           205
one of the desks in the office outside of Robert's cubbyhole. She
was still tucked in the corner, using my parka as a pillow.
  Robert was concerned that Netcom workers coming to work in
the morning might be startled to find a strange woman sleeping
under a desk. I noticed that Andrew had dealt with that problem
by hanging a piece of paper above her head which read: "Do Not
In the first morning light Julia and I stood on a balcony of the
Netcom building opposite Robert's office. Through a cold mist I
could see the early morning commuter traffic was already flowing
on Winchester Boulevard. I'd put my parka on, but the morning
fog still chilled me.
  "Tsutomu, everybody has something to do here but me," she
said. "I feel like a fifth wheel. I shouldn't be here."
  It was true. Through most of the long night Robert had
managed the network, Andrew and I had hacked the filtering
code together, and John Hoffman had set up the monitoring
station. After Julia returned from Soeren's with the driver tape,
she had been left on the sidelines, while the rest of us had been
focused on making this work. She resented being the gofer on
our team.
  I pointed out to her that at the Well she'd been our diplomat,
acting as a liaison with Claudia, whom I was basically ignoring. I
thought back to the previous weekend when Julia decided to
come along on the trip to the Well.
   "When we started this you told me you wanted to be here
because this would be an opportunity to watch and learn," I
reminded her. I could see that she was exhausted and feeling
bad about herself and that something else was troubling her.
But I didn't want to have that conversation now. We had both
been awake for almost twenty hours and we were approaching
                              PROOF                            207
the time the intruder usually became active. We needed to be
back inside because our monitoring gear would require careful
attention, as in the past twelve hours his patterns had changed
and he was no longer coming exclusively through the Raleigh
POP. We had tracked him this far and now we needed to
quickly figure out what the next step would be. My sense of
urgency was growing, and I didn't want this opportunity to slip
away. The longer we waited, the greater the possibility some-
thing would go wrong. I had been hoping he was a creature of
habit, and now I was concerned about losing any more of our
    We stood there for a while longer staring west out toward the
hills, which were barely visible through the fog. It was cold and I
was feeling nauseous, the way I feel sometimes when I've gone
without sleep for too long.
    Finally, to break the silence, I said, "He comes online as early
as 7 A.M., I have to go back inside. I need to make sure we're
    I walked into Robert's office and sat for a while. Having spent
the night with us, he was now back at work doing his regular job.
We watched as the performance monitoring software on his
workstation showed the FDDI network load begin to rise steadi-
ly. There was still a great deal of work to be done, for the infor-
mation that we were now seeing from each POP was being saved
as an undifferentiated glob of data. Without the software that
 Robert had asked Livingston to supply us, we still couldn't break
 it out into individual user sessions. It was as ifwe had been hand-
ed a box containing pieces to several different jigsaw puzzles. We
first needed to sort the separate puzzles from one another, which
 is precisely what the Livingston software would help us do. Only
 then could we piece together what was actually happening in an
 individual session.
    After about an hour Robert suggested we all go downstairs
 and get some breakfast at a deli located on the ground floor of
 Netcorn's building. Andrew had spent all his money, and was
 going to only have a cup of tea, but when I looked through my
 wallet and found six dollars, I gave him most of it so he could
 buy something to eat. Julia had coffee and I bought an Earl
20 8                        PURSUIT
Grey tea in a styrofoam cup, and since Andrew and Robert had
already settled deep into a technical discussion, we walked out-
side to sit in a patio in the courtyard of the building. On the
way outside we laughed about how it was impossible to talk
technology all of the time, the way Robert and Andrew seemed
to. We both agreed a little balance was necessary in life.
    I picked at my empty cup and listened to her. I could hear the
stress in Julia's voice. This was a familiar feeling, every time she
had to go back and deal with John she would become tense and
nervous. But this time she didn't see it.
    ''I'm going to have to get some rest if I'm going to deal with
John this weekend," she said. "If! go up there totally exhausted,
it's just going to be another disaster."
   We spent an hour and a half trying to talk about what was
bothering her, but we weren't getting anywhere and I was feel-
ing increasingly frustrated. We went back upstairs and I con-
tinued to work on our monitoring tools. The software for
untangling the data from the Portmasters had now arrived, but
to be able to use it, I still needed to do a lot of work with my
own tools first.
   Our filter was losing packets and I spent some time fiddling
with the Livingston program to be certain that it would faith-
fully capture any of gkremen's sessions on Netcorn's computers.
   The intruder had returned a little after 10 A.M. and about an
hour later Andrew and Julia went into Robert's office to see what
the intruder was doing. Robert told us that so far today he had
seen gkremen logged in through the Raleigh and Denver POPs.
While they watched over Robert's shoulder, I continued to work
on the software while listening to their comments through the
door to his office.
   The cracker had connected from Netcom to hacktic.nl, the
computer in the Netherlands that is an electronic gathering place
for the computer underground. He was using the account name
martin. Later we would be able to extract a precise videotape of
his activities, but for the moment we had to rely on Robert's
cruder tool that captured characters in a small temporary memo-
ry buffer and displayed them on the screen.
    "My mouse finger is wearing out," Robert said. Every time
                             PROOF                           209
there was a break-in during the last day Robert had been track-
ing it by repeatedly clicking his mouse button and picking up
snippets of the intruder's keystrokes.
   They watched on the screen as the intruder attempted to set up
a talk session with somebody whose user name was jsz. A network
information center database revealed he was located in Israel.The
database also indicated that he was typing from a Silicon
Graphics workstation. The Israeli connection was an interesting
one, for Kevin Mitnick had been rumored to have fled to Israel
while he was a fugitive in California in the mid-1980s. It was
another tantalizing clue.
   The intruder started a program called talk, which divided his
screen in half and allowed him to seewhat he was typing in the top
half of his screen and what jsz responded in the bottom half

[No connection yet]
[Waiting for your party to respond]
[Ringing your party again]
[Waiting for your party to respond]
[Connection established]

martin:    fuck this is lagged

jsz:      hey. OK, just a sec. rim in another window as

martin:   hello

jsz:      hello

martin:   yes, i am lagged to hacktic.

jsz:      AHhh. OK. whats up?
2I 0                        PURSUIT
martin:    can you send me sol & mail stuff?

  After complaining about their very slow connection Martin was
apparently asking his contact in Israel for information. "Sol" prob-
ably referred to Solaris, the version of Unix distributed by Sun
Microsystems, and "mail" might have meant sendmail. Security
weaknesses in mail systems have traditionally been a way to break in
to computers.

jsz:       ok. i sent you sol already.

martin:    i need you to send it again it was corrupted.
           laso can you send me the mail thing now too?

jsz:       ok. yup. let me send it, then,

martin:    ok pls send both again your last pgp msg was
           fucked up.

jsz:       ok. will do let ull again :0

martin:    ok do you want to try the mail thing with me

   There was a long pause at this point. Martin was obviously per-

jsz:       now? no; later i will try myself, perhaps ... you
           want to try it @oki?

martin:    ok can you send me now so I can try it :-)

jsz:       OK. I sent you sol stuff. check it out now,

martin:    yup

jsz:       will dig up and send 8.6.9 thing later.

martin:    hmm. .. i was hoping you would send me the
                             PROOF                           2I I
          mail thing right away so I can get to certain

  Another long pause. They were indeed discussing sendmail; the
current version was 8.6.9.

jsz:      OK. sendmail is sent.

martin:   hold on.

jsz:      check your mail

martin:   i am on the phone too .... ok so you sent me both
          sendmail & sol?
jsz:      yes

martin:   thanks you dont want to try oki now?

jsz:      no

  At this point it seemed that Martin was trying unsuccessfully
to persuade his acquaintance to use his special cracking tools to
attack an Internet gateway computer belonging to Oki Telecom,
the cellular telephone manufacturer.

martin:   ok, are the complete details there in sendmail
          so I can do it wlo your help?

jsz:      take a look, and you'll see yourself. if you
          know how to set up identd, i guess.

martin:   ok, hey are you at labs?

jsz:      not the CS one.

martin:   000 ok well wanna meet online later?

jsz:      yup. dont make this bug useless though :-)

martin:   gimme a break. CERT will do that ina few days : - (

jsz:      HeOOeheh.
2I 2                        PURSUIT
   Jsz was telling Martin not to share this information about a par-
ticular system vulnerability he had just given him. They both knew
as soon as they exploited this weakness the computer security com-
munity would be alerted, and the back door would vanish.

martin:    thanks for the trust. I'll protect it too i want
           to use it just as bad as you!

jsz:       no worries B-)

jsz:       "Give me a place to stand,H said Archimedes,
           "and I will move the earth H :-)
           (just reading through someone's mail, as we
           speak :-)

martin:   :-)

   These guys were using other people's electronic mail the way
most people use the library.
   The session ended, and Robert told everyone to get out of his
crowded office. From the session data, I reviewed what I knew
about our intruder's technique. He obviously believed he was
immune from surveillance. As had been the case at many of the
other computer sites he'd broken in to, he had probably tried to
set up a sniffer at Netcom and found he was unable to monitor
the high-speed FDDI backbone. When he failed, he would have
assumed that because he couldn't install a sniffer, nobody else
could, either. He would have concluded that he had a great secu-
rity advantage by going through Netcom on his first hop into his
pillaging forays on the Internet, where he could not be detected.
He was wrong, for we had accomplished something he had prob-
ably believed wasn't technically possible. In this game we were
playing he had made an incorrect assumption, and he might have
to pay for it.
   I was still sitting in the open space just outside of Robert's
office, mesmerized by the LED scrolling advertising sign on top
of the candy machine when, moments later, he yelled out,
"There's a gkremen session coming from Atlanta."
                               PROOF                            2I 3
   Atlanta! We hadn't seen sessions originating from Atlanta
before. Was the intruder trying to mask his location by entering
Netcom's network from even more locations? I logged back into
looper and patched the Atlanta address into our filter, and imme-
diately the information we were saving turned into a cascade.
Atlanta alone was generating more than nine megabytes of data a
   Worse, I was spooked.
   Until the evening before we had arrived at Netcorn's offices,
their records had shown that the intruder had been connecting
exclusively from their Raleigh dial-up, except when he entered
Netcom via the Internet f~om escape.com or CSN. By Thursday
night, he'd come several times from Denver, and now from
Atlanta; from the logs we saw a brief connection from Chicago,
as well. Worst of all, the new pattern had started almost immedi-
ately after I'd asked Kent, the assistant United States attorney, for
a trap-and-trace order.
   Did he know about the traces? Did he have the ability to tap
the telephone company? Or was he able to eavesdrop on us? If we
were right and our adversary was Kevin Mitnick, he had been
known to illicitly wiretap law enforcement officials to stay one
step ahead of them. It was possible he was just thumbing his nose
at us. If that was the case our task was going to be much harder.
I told myself to wait patiently and not panic, hoping the new pat-
tern was just an anomaly.
   Andrew and Julia returned to watch as the intruder connected
again from Netcom to hacktic.nl, logging in as Martin, with the
password "oki,900."
   He first checked his mail, which had three messages from jsz.
The first was a reply to a query, "Hey, where are you, dude?" It
contained only one line, "okay, back."
   The second and third messages were the text files encrypted
with PGp, or Pretty Good Privacy, the free data encryption pro-
gram. Martin saved them with the file names solsni.asc and send-
mail.asc. Although the file names were intriguing, their contents
were beyond our reach. With a long enough encryption key, PGP
files would even be beyond the decoding powers of the world's
intelligence agencies.
2I 4                        PURSUIT
   Next the intruder typed "w jsz," a command that checked to see
if jsz was still connected to the Hacktic computer, but jsz had van-
ished. Martin now backtracked, disconnecting from Hacktic and
returning to Netcom. He typed, "ftp hacktic.nl" and then again
logged into the Dutch system as Martin. This time he transferred to
gkremen's account the two files jsz had left him from the
Netherlands computer. Finally, he completed the process by down-
loading the files from the gkremen account on the Netcom com-
puter in San Jose to his own personal computer-wherever he was
hidden. After finishing the transfer he immediately deleted the two
files from his account.
   There was a long gause. Was our trespasser thinking? Was he
decrypting and reading his files? Suddenly he connected again to
hacktic.nl and ran a program called Internet Relay Chat, or IRC,
which permits thousands of people allover the world to partici-
pate in hundreds of simultaneous keyboard "chats." It's the CB
radio of the Internet. When IRC requested, "Please enter your
nickname," he typed "marty."
   Marty! Andrew and I had seen "marty" before-it was the
name of the account on the Well where we'd found a stash of
stolen cellular telephone software. With IRC he joined a public
channel called #hack, an open gathering spot for some of the
world's computer underground types. Instantly his screen was
filled with the scurrilous chatter of dozens of anklebiters, much
of it obscene.
    Ignoring the babbling, he sent a private message to jsz. "helo
jsz?"-and then corrected his spelling, "hello jsz?"
    No luck. A message came back, "jsz's away, email me."
   As he was contemplating his next move he was interrupted-
jsz had answered. Acknowledging each other's presence the two
conspirators made secret contact again using the ntalk program.

[No connection yet]
[Connection established]

martin:    hi read that stuff VERY interesting I KNEW
           mastodon would be GOLD!

j sz :     : - ) i knew so too.
                               PROOF                             2I 5
martin:   Hey we need to setup a bd so I can use it too.
          so far I havent FUCKED up one site you let me on
          so history shows. :-) you like history, right?

jsz:       I am history major :)

  They were apparently talking about a computer named
Mastadon where they had apparently found useful information
about a back door ("bd").

jsz:       hehehe. you mean, you want to be on the alias
           too? :-) meanie :0

martin:    hey, thats pretty slick so: when someone con-
           nects to 25 it actually connects back to inetd
           on the remote?

jsz:       back. yes, exactly - that's how it actually
           works (ie identd .. )it might be a nasty bd in
           there :-)

martin:    Heheheh. Why didnt I THNK of it. question: i t
           seems you can dump anything into the que, can it
           execute portd as root or just mail stuff?

jsz:       I am thinking of it. i dont think you can get it
           execute anything as root, but you might trick it
           into running something for you, i will work on
           it later today.

martin:    hmm.. like finger :-)

  As Andrew, Julia, and Robert watched they realized to their
alarm that Martin's Israeli friend had learned about a new securi-
ty hole in sendmail.
   "Hey, they have a new sendmail bug," Andrew said to me
through the doorway. "It has something to do with identd."
   ''I'll turn off the feature right now," I replied. I dropped every-
thing and dialed-in to my computers in San Diego, making sure
that if our intruder tried the new trick on our machines he would
2I 6                       PURSUIT
run into a stone wall. At the same time Robert was on the phone
to John Hoffman, instructing him to do the same thing on all of
Netcorn's computers.
  Julia and Andrew began reading what was being typed on
Robert's screen aloud so I could hear it while I worked on closing
the sendmail hole.

jsz:      eric allman is my hero ;)

martin:   my hero is japboyl

jsz:      finger into markoff's butthole :)

martin:   see markoff is not acting right. a reporter
          doesnt HELP catch someone its not ethical. he is
          the reason why my picture was the front page of
          the new york times

  We had proof! Martin could only be Kevin Mitnick and he was
calling me "Japboy." This was getting personal, but it felt remote
and a little surreal. "That's not very nice," I said.

jsz:      yeah, I think markoff is nigger, he's tired of
          his blaklife, and needs some adventury. He
          should be killed :-) I will send him a package
          from Saddam Hussein, or Collonel Kadaffi, what
          sounds scaries, hussein, or kaddafi?

martin:   nah someone :-) needs to get to nytimes.com and
          create a story about japboy that he is a con-
          victed child molester and get it printed with
          markoff's by line.

jsz:      AHAHAHA., that'd be funny as hell. :-)

martin:   can you imagine the results.

jsz:      tsu will be pissed as hell

martin:   yes, or add to a real markoff story that men-
          tions tsutomu is real dan farmers gay lover. and
          that they secretly meet on queernet.org
                              PROOF                           2I 7
jsz:       for netsex :-) AHAHAHA. That'd be even funnier.

martin:   that would be the hack of the century!

jsz:      hahaha. really :-) markoff will *die* too, tsu
          will find it a matter of honor to buttfuq him :)

  The running commentary continued, but none of us could
believe how childish and inane it all sounded. Rather than sound-
ing like murderers, the pair sounded juvenile.

martin:   hey does 8.6.9 by default connect back to the
          src inetd dor identd?
jsz:      yeah, it ddoes (by default .. ) so does the send-
          mail that casper dick runs :-) dik, even.

martin:   : -) hrmnm ... well obviously we can have it mail
          out shit (example in memo) but to execute code
          is the best technique. :-)

j sz :    you know sendmail techniq       :)

martin:   see I do KNOWtsendmail technique! the trick is
          to do it soon so we get to our targets before
          cert announces the bug.

  At this point I thought back to the first voice mail I'd received
when my caller boasted, "I know sendmail technique." I'd put the
recording on the net, but they obviously knew all about the
break-in to my machines.

jsz:      OK, I will work on it, i think it'd not be hard
          to do it. there aaint many, me thinks, okidoki,
          and some other folks :-)

martin:   well mot, oki, dsys.

  Martin was referring to Motorola, Oki Telecom, and a com-
puter system in Colorado, possible targets to attack.
2I 8                      PURSUIT
jsz:      telnet to 'em and see :-0

martin:   i dont have windows here like you. i would have
          to disconnect from talk then you wanna hold on.
          hold on brb

  At this point Martin exited from the talk program and briefly
checked to see what version of Sendmail the Motorola gateway
computer was running.

xs1% telnet motgate.mot.com 25
Trying .,.
Connected to motgate.mot.com
Escape character is 'A]'.
220 motgate.mot.com. 5.67b/10a - 1.4.4/mot-3.1.1 Sendmail is
ready at Fri, 10 Feb 1995 15:01:15 -0600

500 Command unrecognized
221 motgate.mot.com closing connection
Connection closed by foreign host.

xs1% fg

martin:   no go 5.67b sendrnail i just checked

jsz:      hang on ahh ... thatis IDA sendrnail.

martin:   I suppose it doesnit do the same identd trick.

jsz:      not sure, i run the same sendrnail on netsys

martin:   hey is netsys.com a service which sells shell
          accounts like escape?

  Netsys was a computer system that belonged to a computer
programmer named Len Rose who had been convicted of steal-
ing software from AT&T while serving as a consultant for the
telephone company. He had served a year in jail.
                            PROOF                                 2I 9
jsz:      nope

martin:   how do you connect?

jsz:      you wont be able to connect to netsys from out-
          side even : - )
          try it. nasty firewall by moi :-)

martin: lets see: we can run tap on ramon NOT!

  "Tap," we recognized-it was the program that had been used
to hijack my connection between Osiris and Ariel.

jsz:      ramon is S81, it doesn't support loadable mod-
          ules :) :) . )

martin:   just kidding if   1   could   1   wouldnt violate our

jsz:      I know! Weee, beavis & butthead on MTV (we have
          tv at labs at EE)

martin:   hey they just showed sneakers on t.V. this weel
          old marty whata guy.

jsz:      hmm, we have european channels here, only
          CNN is from the us. [1 saw snelakers a while ago
martin:   hey do we have playnyboy.com yet?

jsz:      no time yet :) will take care of 'em later
          today, i suppose, or tommorow.

martin:   ahh do you have a bd on the sunos.queer box i'll
          do some sniffing if you want.

jsz:      not yet. maddog.queernet.org is the sun, last
          time i checked: what a name, eh ddog.

martin:   normal bd like accessl

  When Andrew mentioned access1 my ears pricked up. That
was the name of a firewall computer that Sun used to protect its
internal corporate network from the unruly Internet. So it had a
back door, too!
220                        PURSUIT
jsz:      AF-your-lastname :-)
          or AF-your-initials :)


jsz:      AFABAI :-) rsh ard.fbi.gov -1 marty csh -fbi

martin:   we have to NEVER let that bd out , so do oyu
          have it installed on maddog?

jsz:      nope, as i said: no time ;-(

martin:   ahhh i thought you would always do a bd for
          later access oh well ...

jsz:      yes will do it sometime this weekend. anyways, i
          am back to coding :-)

martin:   hey we have to go over the procedure again so I
          can start doing it as well. I have some nortes
          but its been a while. too bad your so far away.

jsz:      some nortes?

martin:   notes - sorry.

jsz:      i sent you getpass once :-0
martin:   ahh you did i'll check my other encrypted disk.
          hey you have been REALLY a great help with unix
          stuff. I am going to send you a hole that works
          on EVERY VMS box up to 6.0 by mty friend nmc.

jsz:      wow. that'll be impressive. I wish I knew VMS
          better ;-)

martin:   but NO ONE else has it so its like giving you
          fro PLEASE NEVER share it, ok

jsz:      nmc does, no? fr 1:0) I have none to share it
          with, and I really have no desire to spoil your
          VMS fun :-)

  Nmc was obviously Neill Clift-one more piece of evidence
confirming that Martin was actually Kevin Mitnick.
                             PROOF                            22 I
martin:   great ill pgp it later tonight i am going out
          now. it will work on bguvms :-)

jsz:      Thanks .. I will appreciate it .. Ok, I will be
          coding tonight .. talk to you on tommorow or so.

martin:   I dont even let anyone else know I have it but
          you trust me and I trust you so maybe you can
          use it too in your explorations :-)

jsz:      email me,   ok? thanks!

martin:   ok no problem its actaully the BEST VMS bug I
          have in my toolkit.

jsz:      thanks :-) from remote? : )

martin:   no, not [remote. I dont have a remote bug on VMS
          5.0 and greater.
          but I do for vms 4.7 and lower.

jsz:      cool .. I think bguvms is 6.0,    (not sure, will
          have to check .. still)
          Thanks anyways : - )

  The conversation ended abruptly, apparently because the con-
nection had been dropped, but in Robert's cramped quarters a
mild pandemonium set in. Now there was little doubt about who
the trespasser was. As far as I was aware there was only one com-
puter criminal whose picture had been on the front page of the
New York Times.
  On July 4, 1994, Markoff had written an article in which he
called Mitnick "cyberspace's most wanted." It had described some
of his escapades and noted that he had managed to evade the FBI
and other federal and state law enforcement officials for more
than a year and a half. At the time, I thought that the piece had
made the FBI look inept.
  I called Markoff and told him about the conversation Robert
had just seen, and asked him if there was anyone else's picture
that had appeared on the front page of the Times with one of his
   "The only other person I can think of is Robert Tappan Morris,
and this obviously isn't him," he said.
222                         PURSUIT
   We finally had a face and a set of motives to attach to the
ghostly electronic footprints we'd been following for more than a
month, but there was still a great deal left to be explained.
   Who was jsz? I made some calls, and people I spoke to said
they'd heard of him before. One person said he thought jsz was
working as a subcontractor for a United States semiconductor
company which had a design laboratory in the Middle East.
   It was clear that Mitnick was relying on jsz as a source of
Unix cracking expertise, in exchange for his own knowledge of
DEC's VMS operating system. Talk about honor among
thieves! I also had a hunch that jsz was involved in some way
in the attack on Ariel in San Diego-possibly he'd provided the
tools, or maybe he'd been the one who had actually led the
   We'd also learned another important fact, for while we had
seen that Eric Allman's mail files had been stolen and stored at
the Well, neither Andrew nor I had looked inside them to see
what kind of information they contained. Allman was the
author and maintainer of the sendmail program, and now we
knew that it was likely that jsz had found a detailed discussion
of a new security flaw while reading through Allman's mail after
breaking into mastodon.cs.berkeley.edu, the computer where it
was stored.
   Mitnick and jsz were systematically trawling the Internet, and
it appeared they were specificallytargeting the computers of secu-
rity experts to rifle through their mail. With the techniques they
had pilfered, they subsequently attacked the computers of corpo-
rations like Apple, Motorola, Oki, and Qualcomm.
   At 2: 11 P.M. Mitnick connected from Denver through Netcorn
to escape.com. From my post next to the soft drink machine I
could hear the other members of our team laughing as they
watched him copy a file called girls.gif into a directory belonging
to jsz on Escape. He then looked through Markoff's mail file and
scanned the subject headings, stopping only to read one person-
al note that had been sent to Markoff by a friend.
   Several minutes later he found his friend jsz again for an
impromptu chat:

Message from Talk_Daemon@escape.com at 17:20
connection requested by jsz@ramon.bgu.ac.il
                            PROOF                             223
respond with: ntalk jsz@ramon.bgu.ac.il

martin:   hi there

jsz:      hi

martin:   what bd are you plotting :-(
          :-) i mean
jsz:      :-)
          we'll see after it's done!
          it'll execute portd itself.

martin:   I can hardly wait; is it sexy?

jsz:      yah, dan farmer would fall for it :)

  In their dreams maybe. Don't these guys have a life?

martin:   hehehh. ok ill let you get back to it i am going
          out now to at lunch and work on finding a real

jsz:      send me pizza . )
          (kosher tho) :) ok.
martin:   with ham?
jsz:      good luck with your searches.

martin:   Can you sugest any good reading books on sysadm
          on unix boxes?

jsz:      sure: read cyberpunk :)

martin:   ya yay ya

  It looked like Mitnick was using Cyberpunk as his resume.

martin:   i moved some file into escape:-jsz/marty ok

jsz:      hey, you can use -jsz/.elm/.4_m dir for your-
          self .. if you want .. I will make it world-write-
          able, but not readable, so you'd have to know
          the EXACT path to it.

martin:   ok will move them later, ya since posse fucks
          with you maybe I should.
    224                         PURSUIT
    jsz:      hehehe. ok.

    martin:   they will fuck with me by mistake and they dont
              want to piss me off :-)

    j sz :    AHAHAHAHAHAa. B-)

    martin:   or their phones wont have dialtone!

    jsz:      go for it : ) naah.

    martin:   ok i will talk withn you later.

    jsz:      Ok ,   talk you later today ..

    martin:   bye

    jsz:      bye!

       Nice guys, I thought. Is this how they spend every waking
    moment? I began to recall what I knew about the Posse, the gang
    that Jim Settle, the former FBI agent, believed was responsible
    for breaking-in to my computers. What did we have here,
    internecine war in the cyber underground? My reverie was
    interrupted by Robert, who had pushed back from his desk and
    stood up.
       "I have to get some sleep, and you have to go, because I can't
    leave you here," he said determinedly. Robert had now been up
    for more than thirty hours, and three in the afternoon was the
    customary time the trespasser would generally take time out. He
    assured me that his pager was working and that if Mitnick
    returned, we would be alerted. I realized he was right, and in any
    case, for the first time in days I felt comfortable with our posi-
    tion. The monitoring system was now working, we were fairly
    confident that we knew who our target was, and we needed to get
    some food and rest, for we were going to have to be prepared to
    deal with Mitnick later in the evening. From here on it would be
    necessary to work with the FBI and the telephone companies to
    pinpoint precisely where he was physically located. We agreed to
    meet again at 8 P.M.
       Out on the street Julia, Andrew, and I were in that zombie-like
                               PROOF                             225
state that comes when you go beyond lack of sleep, but I couldn't
take time to rest, for there was much to be organized if we were
going to see any progress that night.
   I needed to make phone calls, and so we drove to a Hobie's
restaurant located a couple of blocks away in a shopping center.
Hobie's is a distinctly California-style health food chain special-
izing in serving breakfast all day-it's the kind of place where
sprouts seem to come on everything you order. Inside we spread
our gear out on our table. We were trying to be circumspect, but
with a cell phone, beepers, and a RadioMail terminal, we were
probably pretty noticeable, which gave us all a tinge of paranoia.
Julia thought people around us were listening to our conversation
while trying to appear as if they weren't.
   Finally the waitress walked over, took a look at the table littered
with electronics, and said, "You guys look like you've been
   "Still working," I answered.
   I asked her about the difference between two of their veggie
burgers and she launched in to a detailed and entertaining tech-
nical dissertation about the distinction between their Soy Burger
and their Garden Burger. One of them, it turned out, came
with mozzarella cheese. It tasted better, but it was higher in fat.
   The other house speciality was fruit smoothies, and we each
ordered one. When they came the waitress offered to spray whipped
cream on them from a can. In my mind, that basically expresses the
true nature of the California approach to health food restaurants.
They're healthy, at least ostensibly, but actually they tend to be
places you can go and not feel guilty about eating junk food.
   After we finished ordering I walked outside to the pay phone and
called Kent Walker. I'd already talked to him several times during
the day to check the status of the trap-and-trace orders. I told him
we were almost certain it was Mitnick we were after, and asked
again if he had had any luck in getting the Denver trap-and-trace.
   "Tsutomu, would you like the good news or the bad news
first?" he said, sounding clearly unhappy.
   ''I'll take the bad news," I replied.
   "The Denver assistant United States attorney called the Los
Angeles FBI and they told him not to do anything on this," he
22 6                       PURSUIT
   "Can I infer that I've stumbled into a turf war?" I asked him.
   Kent didn't respond to that, but he didn't need to. "The good
news," he said, "is that we have a trap-and-trace in Raleigh that
should have been active as of 5 P.M. East Coast time."
   Getting a warrant for Raleigh meant that the next time
Mitnick connected to Netcorn from their Raleigh POP, the
phone company would be able to determine where the caIl was
coming from.
   Still I wasn't ready to give up on the Denver order.
   "Can they actually do this? Isn't this obstructing justice or
something like that?" I asked him. "The most recent activity has
been from Denver, and it would be really good if we could get
trap-and-trace there as weIl for the weekend."
   "Hey look, Tsutomu, it's four-thirty now in Denver," he said.
"it's almost the end of the day."
   "Well, you still have a half an hour," I pressed, still hoping.
   After a long pause he said, 'TIl try, but don't count on any-
   Kent gave me Levord Burns' SkyPager number and told me
that the next time we had a Raleigh connection, Burns would
help us get the trace for it from the phone company. When I
groaned and asked why I couldn't go to the phone company
directly, he gave me some telephone numbers of phone compa-
ny people, but told me I should try to reach the FBI contact
   I thanked him and hung up. Just because the L. A. FBI had
been chasing Mitnick for more than two years, that wasn't a
reason why we shouldn't be able to conduct our own investiga-
tion as well. It was frustrating, but I was glad that Kent was
with us.
   We left Hobie's, and Andrew drove off to pick up some sup-
plies, including fresh clothes, for he hadn't had time to go back
to Pei's the night before to gather his belongings. Julia and I
walked across the Hobie's parking lot to a field which was fuIl
of wet green grass and flowering yellow mustard. The ground
was wet with rain, but it was as close as we could get to nature
in the midst of San Jose's suburban concrete sprawl.
   It was still several hours before Robert was returning, so we
climbed into Julia's car and dragged the sleeping bags over us in
the front seats. A while later Andrew drove up and parked near
                              PROOF                            227
us. He left the +4 Jeep's engine running to stay warm and
reclined the front seat and and fell asleep.
   It was almost dark, and we lookedover at Andrew's car and saw
that a good samaritan was peering in his window wondering if
the person who appeared comatose with the engine of his car
running was in need of assistance. We assured her he was okay, he
just been up all night and was resting.
   Having had lunch, Julia seemed a little more energized, but
was still very tense. It was almost time for her to leave for the
weekend, and she didn't feel she was prepared to deal with John.
It would have been difficult enough even under ideal circum-
stances, but now she was exhausted and fearful that she wouldn't
be able to hold her own.
   Given her current state I was also nervous about her being
with John for the weekend. At this point in the investigation
the last thing I wanted to have to deal with were Julia's conflicts.
I was feeling badly overloaded and I was juggling a series of
balls trying to keep this operation going forward: technical,
legal, and political. I was feeling that I couldn't deal with any
more stress.
   We talked for a while about whether she really should go
away for the weekend. After a while I found myself growing
frustrated because it seemed to me that Julia was trying to put
me at ease rather than facing her own internal turmoil and dis-
   "Look," I said. "Whether I'm concerned about your going is
my problem, but whether you should go or not is something you
need to decide for yourself."
   She considered this for a moment.
   "Tsutomu, having come this far I'd like to see the investigation
all the way through," she said.
   "I can't tell you that nothing is going to happen while you're
away," I replied. I don't know what's next, only that if we don't
move quickly, we stand a good chance of losing him."
   "If I leave now am I going to miss the end game?" she asked.
   "I'm hoping to have trap-and-trace information from Raleigh
tonight," I said, "and as soon as we have a solid lead, I'm mov-
ing our base of operations. If he's coming from Denver, I'm
heading in that direction; if it's Raleigh, then that's where I'm
gomg. "
228                         PURSUIT
  I told her that she could try to catch up, but we couldn't wait
for her.
  She was feeling uncertain about what she should do because
she had made a promise to John and felt she had to keep it.
  I found myself getting even more frustrated.
  "You want to have the best of both worlds and you can't do
that," I told her. "Ar some point you're going to have to make a
  It got later and later while she wrestled with whether she was
going to stay or go. Several times she called John on the cell
phone to tell him she was going to be late. It was 7:30 P.M. when
she finally decided to leave. I got out of the Mazda and we trans-
ferred my ski gear into the +4 Jeep. Julia said that she would
come and join me when she got back and I told her she could get
in touch with Andrew to track me down.
  I had now been up for almost thirty-two hours. Andrew and I
drove back to Netcom and waited for a while, dozing, and then
we went back upstairs to continue our watch.

Back inside, under the unrelenting soft hum of the fluorescent
lights, we looked over the logs and saw that Mitnick had been
gone for more than an hour. His last session, which had ended at
6:58 P.M., had come through Denver.
   I was worried. We had the ability to trace him in only one city
out of dozens of possibilities, and he now seemed to be avoiding it.
I was hoping that it was only the technical problems that Netcom
was having in Raleigh that was causing him to reroute his calls, but
I couldn't be sure that he wasn't already tracking us as well.
   I was also bleary with exhaustion, and though it would have
been an incredible luxury to go off and find a hotel to crash in, I
knew that tonight might be our best and only opportunity to get
a trace on him. Since I was very young I've always had the abili-
ty to stay awake for long stretches by putting myself in a zone and
just focusing on the problem. But as I sat in front of my com-
puter that Friday and tried to make sense of the data we had col-
lected during the day I could see that this was a capacity that was
deserting me with age.
   As I sat with Andrew getting our software tools in order and
                              PROOF                            229
waiting for Mitnick to return I realized that Julia's disappearance
had left me with a sense of relief. I was surprised, but I hadn't
realized how distracted I'd been. Now I felt I could finally focus
on the hunt. Although it was Friday evening, there was more
than the usual activity at Netcom as a handful of telephone
installers moved through the offices, replacing telephones on
each desk. I got up and walked into Robert's office at one point
and realized there was a large, fresh pyramid of Coke cans sitting
on his desk. We had been here a long time.
   I went back to looking over the information we'd logged, when
at 10:44 Mitnick logged on. And he was calling from Raleigh!
   "Andrew, why don't you see if you can wake up Levord?"
Now it was time to see if the FBI could make its own contri-
bution. Levord was asleep at his home in Fairfax, Virginia, a
Washington suburb, but he said he would work on it. Fifteen
minutes later he called back. Andrew spoke to him briefly and
then leaned his head out the office door and said, "He says that
the GTE guys told him the phone number we gave them does-
n't exist."
   I looked at my notes and dialed the number myself. Through
the earpiece I heard the familiar mating call of a high-speed
modem. I walked into Robert's office and took the phone from
   "Hey, it works for me, would you like to hear the modem car-
rier tone?" I said angrily. "What is going on with those bozos?"
   This squared with everything I'd heard about GTE. Just our
luck, I thought to myself. Levord sounded as if he was still
more asleep than awake, but promised to ask them to check
   Unlike some of Mitnick's sessions, this one was a long one, last-
ing nearly thirty-five minutes.
   Shortly after Mitnick logged off, Levord called back and said,
"He's gone, they didn't get a trace."
   "Yeah, well, your guys had a half an hour."
   Levord didn't sound that upset. "If you see him on again, call
me back," he answered. "They have the tracing equipment in
pIace now."
   "My understanding was that they were supposed to have it
working eight hours ago!" I told him.
230                         PURSUIT
   "They didn't seem to have heard anything about it," he replied.
   It was now 11:20, but fortunately we had to wait for only a few
minutes. Robert's pager buzzed again, and sure enough, Mitnick
was back as gkremen, once again from Raleigh.
   "Call Levord back again and tell him to get these guys going,
and get them to trace where the call is coming from this time," I
said to Andrew.
   He made the call and we waited again.
   Thirty minutes later the phone rang.
   Special agent Burns reported that they had a successful trace.
They had a phone number, which was assigned to Centel, a local
cellular phone company recently acquired by Sprint Cellular.
Beyond that he wouldn't tell us anything. But we had a phone
number, and that might lead us to a physical address! It looked as
if the FBI had been right: Mitnick was making his data calls via cel-
lular phone. Levord and I agreed that in the morning he would
contact Sprint and arrange to trace the call through their switch.
   Mitnick was still electronically prowling around the Netcom
computers. As Robert watched he connected to one of the com-
pany's server computers called Netcomsv. This was a machine
that ran special services available for all users such as the Usenet
computer conferencing system, and we discovered that he had
installed a back door. He logged in as root and used a password
".neill."-he was still obsessing about Neill Clift-and looked
around for a while and then left. Robert was furious. He got on
the phone to John Hoffman and made sure that the backdoor
was closed immediately
   We continued to follow the sessions from the evening and
watched a remarkable attack unfold. From Netcom Mitnick had
connected to CSN and made himself root at a little after 11:30
P.M. He was still fiddling around with the operating system files of
one of their main computers, attempting to install and conceal the
system we had seen him build the night before. After about a half
an hour he managed to reinstall NIT successfully and then he
rebooted-restarted-the computer to make it run his program.
He had done it right under the noses of the CSN administrators!
   As we watched the session with our own monitoring software
Andrew turned to me and said, ''A lot of gal!!"
                             PROOF                          23I
  As a system administrator, Robert couldn't believe what he was
seeing. "I want a videotape of that," he told me.
  We watched for a few minutes more, but it had become obvi-
ous that none of us were going to be able to fight off sleep any
longer. It had been thirty-nine hours since I'd woken up in San
Francisco on Thursday morning, and exhaustion had taken its
   I hunted through the phone book and found a nearby
Residence Inn. I reserved two rooms, and Andrew and I drove
about five kilometers through San Jose's deserted streets. It was
three in the morning by the time we'd checked in and fallen
I had certainly been aware of Kevin Mitnick long before the
traces of oki.tar.Z on Ariel made him a suspect in my break-in.
He had achieved legendary status in the computer underground
over a fifteen-year period stretching back as far as 1980, and he'd
had numerous run-ins with local, state, and federal law enforce-
ment officials and been sent to jail several times.
   My first run-in with him took place during the summer of
1991, when he attempted to "social engineer" a piece of com-
puter security information from me over the telephone. "Social
engineering" is a tactic used by people in the computer under-
ground to access a computer by talking unsuspecting computer
system administrators and telephone company employees out of
valuable information. They rely on people wanting to be helpful.
When someone calls and says that they are a new employee in the
company, or somebody in another division who has misplaced a
password, or someone with a legitimate need for temporary net-
work access to a computer, a person's natural inclination is to give
him the information.
   Kevin's call carne a few months after I discovered a fairly glar-
ing security loophole in Digital Equipment Corporation's
ULTRIX operating system. You could get root on a DEC work-
station by sending the computer an electronic mail message to a
magic address and then typing a few commands. This bug was
what software designers call a "buffer overflow condition," and
                              KEVIN                           233
Robert Tappan Morris's worm program exploited a similar flaw in
a network service supplied with the Unix operating system. The
software was expecting a character string of no more than a cer-
tain length, and when it got a longer one, rather than handling it
gracefully, the program could be made to alter its behavior in a
particularly nasty and odd way that had the consequence of
granting the user all system privileges.
   I described the bug in a message to CERT. In principle, CERT
is supposed to serve as a clearinghouse for information on com-
puter vulnerabilities so that the people responsible for adminis-
tering computer networks can learn about and patch them before
the computer underground can exploit them. The reality is that
rather than make such information freely available so that securi-
ty loopholes will be attended to, CERT has instead attempted to
control its spread as much as possible. It will never publicize the
names of organizations that have suffered a break-in, arguing that
is the only way it can obtain cooperation. It also tends to produce
advisories so general they are not very helpful.
   A few months after I reported the ULTRIX bug, CERT pro-
duced an advisory that described it in so sanitized a fashion that
the report didn't provide enough information to enable someone
to reproduce the error. Brosl and I had moved from Los Alamos
to San Diego by this time, but I had flown back to Los Alamos
to spend a week at the Center for Nonlinear Studies. One morn-
ing I heard from my secretary in San Diego that she was getting
repeated phone calls from someone at Sun Microsystems who
said it was urgent that he reach me. Several hours later I was sit-
ting in a borrowed office when the phone rang.
   "Hello, this is Brian Reid. I'm a field specialist for Sun
Microsystems in Las Vegas." The caller was speaking smoothly
and rapidly. He told me that he had seen the CERT advisory and
that he was now at a customer site and needed more information.
''I'm not able to recreate the flaw," he explained.
    I was immediately on guard. I knew of one Brian Reid, but he
worked at DEC, not Sun. This made no sense. First of all, why
would someone from Sun Microsystems, at a customer site, be so
anxious to get technical information about a security flaw in one
of his competitor's computers?
    "How can I verify who you are?" I asked.
234                          PURSUIT
    "That's not a problem," he replied. "Just call this directory
 number at Sun and they will confirm that I work for them."
    He gave me the Sun number, as well as a number in the 702
 area code where I could reach him, and then hung up. I called my
 friend Jimmy McClary, a Computer Systems Security officer for
 Los Alamos National Laboratory, and told him about the call. He
 came downstairs and sat with me while I dialed the Sun number
 the caller had given me. I asked the operator for an employee
 named Brian Reid, and was told there was no such person work-
 ing for Sun. I hung up and was chatting with Jimmy about what
 to do about the caller, when my phone rang again.
    This time a much less professional-sounding voice identified
 itself as a co-worker of Brian Reid at Sun, and said he was also
.attempting to obtain the information that Mr. Reid had
    "Why don't you give me your address, and I'll drop it in the
 mail on a floppy disk?" I suggested. This seemed to startle my sec-
 ond caller, who began to "urn' and "ah." Finally he came up with
 an address that sounded as if it had been made up on the fly, and
 hung up abruptly.
    I tried the 702 number the first caller had left, and I got the
 whistle of a computer modem. The 702 area code covers all of
 Nevada, so I gave the number and the address to Jimmy, who
 went off to phone the Department of Energy security officials
 about it. Later I learned that the security people had traced the
 number I had been given to a pay phone on the University of
 Nevada at Las Vegas campus. Some of these phones cannot
 receive phone calls, but instead have a modem to communicate
 billing information and diagnostics.
    Several weeks later I was speaking with Markoff and when I
 began describing the call from someone claiming to be Brian
 Reid, Markoff started laughing.
    "What did I say?" I asked him.
    "There is only one person who would use the name Brian Reid
 while he was trying to social-engineer you," he replied.
    Markoff, who had been researching Kevin Mitnick for his book
 called Cyberpunk, explained that in 1987 and 1988, Kevin and a
 friend, Lenny DiCicco, had fought a pitched electronic battle
  against the real Brian Reid, a scientist at DEC's Palo Alto research
                               KEVIN                            23 5
laboratory. Mitnick had become obsessed with obtaining a
copy of the source code to DEC's VMS minicomputer operat-
ing system, and was trying to do so by gaining entry to the
company's corporate computer network, known as Easynet.
The computers at DEC's Palo Alto laboratory looked the most
vulnerable, so every night with remarkable persistence Mitnick
and DiCicco would launch their modem attacks from a small
Calabasas, California, company where DiCicco had a comput-
er support job. Although Reid discovered the attacks almost
immediately, he didn't know where they were coming from,
nor did the local police or FBI, because Mitnick was manipu-
lating the telephone network's switches to disguise the source
of the modem calls.
   The FBI can easily serve warrants and get trap-and-trace infor-
mation from telephone companies, but few of its agents know
how to interpret the data they provide. If the bad guy is actually
holed up at the address that corresponds to the telephone num-
ber, they're set. But if the criminal has electronically broken-in to
the telephone company's local switch and scrambled the routing
tables, they're lost. Kevin had easily frustrated their best attempts
at tracking him through the telephone network using wiretaps
and traces. He would routinely use two computer terminals each
night-one for his forays into DEC's computers, the other as a
lookout to scan the telephone company computers to see if his
trackers were getting close. At one point, a team of law enforce-
ment and telephone security agents thought they had tracked
him down, only to find that Mitnick had diverted the telephone
lines so as to lead his pursuers not to his hideout in Calabasas, but
to an apartment in Malibu.
   Mitnick, it seemed, was a tough accomplice, for even as they
had been working together he had been harassing DiCicco by
making fake calls to DiCicco's employer, claiming to be a gov-
ernment agent and saying that DiCicco was in trouble with the
Internal Revenue Service. The frustrated DiCicco confessed to
his boss, who notified DEC and the FBI, and Mitnick soon
wound up in federal court in Los Angeles. Although DEC
claimed that he had stolen software worth several million dollars,
and had cost DEC almost $200,000 in time spent trying to keep
 him out of their computers, Kevin pleaded guilty to one count of
236                         PURSUIT
computer fraud and one count of possessing illegal long-distance
access codes.
   It was the fifth time that Mitnick had been apprehended for
a computer crime, and the case attracted nationwide atten-
tion because, in an unusual plea bargain, he agreed to spend one
year in prison and six months in a counseling program for his
computer addiction. It was a strange defense tactic, but a feder-
al judge, after initially balking, bought the idea that there was
some sort of psychological parallel between the obsession
Mitnick had for breaking-in to computer systems and an addict's
craving for drugs.

Kevin David Mitnick reached adolescence in suburban Los
Angeles in the late 1970s, the same time the personal computer
industry was exploding beyond its hobbyist roots. His parents
were divorced, and in a lower-middle-class environment in which
he was largely a loner and an underachiever, he was seduced by
the power he could gain over the telephone network. The under-
ground culture of phone phreaks had already flourished for more
than a decade, but was now in the middle of a transition from the
analog to the digital world. Using a personal computer and
modem it became possible to commandeer a phone company's
digital central office switch by dialing in remotely, and Mitnick
became adept at doing so. Mastery of a local telephone company
switch offered more than just free calls: it opened a window into
the lives of other people; to eavesdrop on the rich and powerful,
or on his own enemies.
  Mitnick soon fell in with an informal phone phreak gang that
met irregularly in a pizza parlor in Hollywood. Much of what
they did fell into the category of pranks, like taking over directo-
ry assistance and answering operator calls by saying, "Yes, that
number is eight-seven-five-zero and a half Do you know how to
dial the half, rna'am?" or changing the class of service on some-
one's home phone to that of payphone, so that whenever they
picked up the receiver a recorded voice asked them to deposit
twenty cents. But the group seemed to have a mean streak as well.
One of its members destroyed files of a San Francisco-based
computer time-sharing company, a crime that went unsolved for

                                   KEVIN                            23 7
    more than a year until a break-in at a Los Angeles telephone com-
    pany switching center led police to the gang.
       That break-in occurred over Memorial Day weekend in 1981,
    when Mitnick and two friends decided to physically enter Pacific
    Bell's COSMOS phone center in downtown Los Angeles. COS-
    MOS, or Computer System for Mainframe Operations, was a
    database used by many of the nation's phone companies for con-
    trolling the phone system's basic recordkeeping functions. The
    group talked their way past a security guard and ultimately found
    the room where the COSMOS system was located. Once inside
    they took lists of computer passwords, including the combina-
    tions to the door locks at nine Pacific Bell central offices, and a
    series of operating manuals for the COSMOS system. To facili-
    tate later social engineering they planted their pseudonyms and
    phone numbers in a rolodex sitting on one of the desks in the
    room. As a flourish one of the fake names they used was "John
    Draper," who was an actual computer programmer also known as
    the legendary phone phreak, Captain Crunch. The phone num-
    bers were actually misrouted numbers that would ring at a coffee
    shop pay phone in Van Nuys.
       The crime was far from perfect, however. A telephone company
    manager soon discovered the phony numbers and reported them
    to the local police, who started an investigation. The case was actu-
    ally solved when a jilted girlfriend of one of the gang went to the
    police, and Mitnick and his friends were soon arrested. The group
    was charged with destroying data over a computer network and
    with stealing operator's manuals from the telephone company.
    Mitnick, seventeen years old at the time, was relatively lucky, and
    was sentenced to spend only three months in the Los Angeles
    Juvenile Detention Center, followed by a year's probation.
       A run-in with the police might have persuaded most bright
    kids to explore the many legal ways to have computer adventures,
    but Mitnick appeared to be obsessed by some twisted vision.
    Rather than developing his computer skills in creative and pro-
    ductive ways, he seemed interested only in learning enough
    short-cuts for computer break-ins and dirty tricks to continue to
    play out a fantasy that led to collision after collision with the
    police throughout the 1980s. He obviously loved the attention
    and the mystique his growing notoriety was bringing. Early on,
238                        PURSUIT
after seeing the 1975 Robert Redford movie Three Days of the
Condor, he had adopted Condor as his nom de guerre. In the film
Redford plays the role of a hunted CIA researcher who uses his
experience as an Army signal corpsman to manipulate the phone
system and avoid capture. Mitnick seemed to view himself as the
same kind of daring man on the run from the law.
   His next arrest was in 1983 by campus police at the University
of Southern California, where he had gotten into minor trouble
a few years earlier when he was caught using a university com-
puter to gain illegal access to the ARPAnet. This time he was dis-
covered sitting at a computer in a campus terminal room, break-
ing into a Pentagon computer over the ARPAnet, and was sen-
tenced to six months at the California Youth Authority's Karl
Holton Training School, a juvenile prison in Stockton,
California. Afrer he was released, he obtained the license plate X
HACKER for his Nissan but he was still very much in the com-
puter break-in business. Several years later he went underground
for more than a year after being accused of tampering with a
TRW credit reference computer; an arrest warrant was issued, but
it later vanished from police records without explanation.
   By 1987, Mitnick seemed to be making an effort to pull his
life together, and he began living with a woman who was tak-
ing a computer class with him at a local vocational school. After
a while, however, his obsession drew him back, and this time his
use of illegal telephone credit card numbers led police investi-
gators to the apartment he was sharing with his girlfriend in
Thousand Oaks, California. He was convicted of stealing soft-
ware from the Santa Cruz Operation, a California software
company, and in December 1987, he was sentenced to thirty-
six months probation. That brush with the law, and the result-
ing wrist slap, seemed to only increase his sense of omnipo-
   In the summer of 1988 Markoff obtained a copy of a confiden-
tial Pacific Bell memorandum from a teenage computer hacker.
The phone company had no idea how it had been leaked, but
confirmed that it was authentic. The memo, written the year
before, concluded that "the number of individuals capable of
entering Pacific Bell operating systems is growing" and that
"computer hackers are becoming more sophisticated in their
attacks." As a result, the document acknowledged, personal com-
                              KEVIN                            239
puter users could illegally connect their machines to the phone
network and with the proper commands could eavesdrop, add
calls to sorneone's bill, alter or destroy data, intercept facsimile
documents being transmitted, have all calls to a particular num-
ber automatically forwarded to another number, or make some-
one's line appear permanently busy. In one of the cases cited, a
group of teenage computer hobbyists was able to pull such stunts
as "monitor each other's lines for fun" and "seize another person's
dial tone and make calls appear on their bill." One of the crack-
ers used his knowledge to disconnect and tie up the telephone
services of people he didn't like. In addition, "he would add sev-
eral custom-calling features to their lines to create larger bills."
   The leaked memo was described in a July 1988 front-page arti-
cle in the New YOrk Times written by Markoff and Andrew
Pollack. Although he did not know it at the time, Markoff later
learned that Mitnick had been the source of the document.
Mitnick, whose technical tools included amateur radio, had
heard about the memo from a fellow radio operator. Placing a call
to the secretary of the author, a security executive at the tele-
phone company, he masqueraded as another Pacific Bell execu-
tive and asked her to fax him a copy of the memo. What the sec-
retary didn't know was that Mitnick had rerouted the telephone
number so that the memo, instead of being received by a Pacific
Bell fax machine, was soon scrolling off a fax machine in the
office of a friend. The friend had even programmed the machine
so that the secretary received confirmation that the document
had reached the correct fax number.
   While the Southern California press would soon be referring to
Mitnick as the "Dark Side Hacker" and "the John Dillinger of the
computer underground," he was really more of a con man or a
grifter than a hacker in the true sense of the word. Before the
 1983 movie war Games, in which Matthew Broderick portrayed
a young man with some of Kevin Mitnick's traits, the word
"hacker" had been used to refer to a computer culture that had
emerged at MIT in the late 1950s. The culture was made up of
mostly young men who were obsessed with complex systems as
an end in their own right, a culture that was based on the princi-
ples of the open sharing of software and hardware designs with
friends and of the creation of clever "hacks"-ingenious pro-
grams that pushed the state of the computing art.
240                        PURSUIT
  The true hackers were people like Richard Stallman, who as an
MIT student during the 1970s wrote EMACS, a programmer's
editing tool. EMACS gave programmers a way to repeatedly
revise programs to approach a perfect state, and versions of it
are still in wide use by many, if not most, of the nation's best
programmers today. But after war Games became a blockbuster
in 1983, the popular definition of "hacker" became a teenager
with a modem who was brazen enough to dial into a Pentagon
computer. The true hacker community has been attempting to
reclaim the original spirit and meaning of the word ever since,
but to little avail. A particularly discouraging incident occurred
in 1987 when a small annual gathering known as the Hacker's
Conference invited a CBS news crew to attend its meeting in
the hills above Silicon Valley. The Hacker's Conference is a
low-key event, and perhaps the only professional conference
that offers attendees a second full dinner, at midnight, to take
account of the nocturnal habits of the hackers. Unfortunately
the CBS reporter wasn't about to let the mundane truth get in
the way of a good story. He began his broadcast with the
alarmist warning that he had visited the encampment of a
guerrilla army set to undermine the country's security in a new
kind of information warfare.
  The world he was describing had little to do with true hackers,
but was home to a growing number of people like Kevin Mitnick.

After he finished his prison time and his halfway-house counsel-
ing sentence for the 1989 DEC conviction, Mitnick moved to
Las Vegas and took a low-level computer programming position
for a mailing list company. His mother had moved there, as had
a woman who called herself Susan Thunder who had been part of
Mitnick's phone phreak gang in the early 1980s, and with whom
he now became reaquainted. It was during this period that he
tried to "social engineer" me over the phone.
   In early 1992 Mitnick moved back to the San Fernando Valley
after his half-brother died of an apparent heroin overdose. He
briefly worked for his father in construction, but then took a job
he found through a friend of his father's at the Tel Tec Detective
Agency. Soon after he began, someone was discovered illegally
                              KEVIN                           241
using a commercial database system on the agency's behalf, and
Kevin was once again the subject of an FBI investigation. In
September the Bureau searched his apartment, as well as the home
and workplace of another member of the original phone phreak
gang. Two months later a federal judge issued a warrant for
Mitnick's arrest for having violated the terms of his 1989 proba-
tion. There were two charges: illegally accessing a phone company
computer, and associatingwith one of the people with whom he'd
originally been arrested in 1981. His friends claimed Mitnick had
been set up by the detective firm; whatever the truth, when the FBI
came to arrest him, Kevin Mitnick had vanished.
   In late 1992 someone called the California Department of
Motor Vehicles office in Sacramento, and using a valid law
enforcement requester code, attempted to have driver's license
photographs of a police informer faxed to a number in Studio
City, near Los Angeles. Smelling fraud, DMV security officers
checked the number and discovered that it was located at a
Kinko's copy shop, which they staked out before faxing the pho-
tographs. Somehow, the spotters didn't see their quarry until he
was going out the door of the store. They started after him, but
he outran them across the parking lot and disappeared around
the corner, dropping the documents as he fled. The agents later
determined that the papers were covered with Kevin Mitnick's
fingerprints. His escape, subsequently reported in the newspa-
pers, made the authorities look like bumblers who were no match
for a brilliant and elusive cyberthief.
   Mitnick's disappearance sent agents of the Los Angeles FBI
down a series of blind alleys. During his time on the lam
Mitnick used his social engineering skills to resume harassing
Neill Clift, a British computer researcher from whom he had
stolen information while he was battling DEC several years ear-
lier. In 1987 one of the richest treasure troves for Mitnick had
been in reading the electronic mail of DEC's security experts.
There he had found private messages detailing security flaws
that had been discovered in the company's VMS operating sys-
tem. Clift, who explored the weaknesses in the system as some-
thing of a hobby, informed DEC of his findings so the compa-
ny could fix the problems.
   Now Mitnick once again began breaking into computers that
242                         PURSUIT
Clift used. In a series of long international telephone calls,
Mitnick, who has an actor's talent for altering his voice, also con-
vinced Clift that he was an employee of DEC, interested in
obtaining details of new security flaws that Clift had found in the
latest release of the VMS system. At Clift's request, Mitnick sup-
plied him with DEC technical manuals that he believed could
only have come directly from the company. The two men then
agreed to set up an exchange of data, encrypting it with PGP. Clift
sent Mitnick a detailed accounting of the latest security flaws he
had found, but in a later phone conversation he became suspi-
cious and realized that he had been tricked. Without letting on to
Mitnick, Clift contacted the FBI, which attempted for weeks to
trace the calls without success. Around the same time Clift was
contacted by the Finnish Bureau of Economic Crimes, which sus-
pected that Mitnick had stolen software source code from Nokia,
a Finnish cell phone maker with a factory in California.
   After receiving a mysterious phone call requesting a Nokia
technical manual, the company mailed it to the specified address,
a California motel, but alerted the FBI. Agents staked out the
motel, only to find that someone had called the front desk and
had the package forwarded to a second motel, and the trap failed.
Several weeks later, Mitnick somehow discovered the FBI's tele-
phone-tracing attempts and outraged, phoned Clift, called him a
"stool pigeon," and again disappeared.
   In March 1994 the FBI was publicly embarrassed after they
showed up at a gathering of the civil liberties and computers
crowd at an annual event known as the Computers, Freedom,
and Privacy Conference, and arrested an unfortunate attendee
whose only crime was the mistake of registering at the conference
under one of Mitnick's aliases. He was nabbed in his hotel room
 in his underwear, and although he and his roommates protested
that he was not Mitnick, the man was handcuffed, and taken to
the local FBI office. His fingerprints were checked, and about
 thirty minutes later word came back that the arrested man was
 not the fugitive. The FBI had to take him back to his hotel, apol-
 ogizing profusely.
   At about the same time Markoff received a call from
 Qualcomm, a San Diego firm developing a new digital cellular
 telephone technology known as code division multiple access, or
                             KEVIN                           243
CDMA. This technology is particularly valuable because it will
allow cellular telephone service providers to pack many times the
number of calls into the same amount of radio frequency spec-
trum. Qualcomm was in the process of setting up a joint manu-
facturing plant in San Diego with the Sony Corporation to build
the new hand-held digital telephones that will employ the
CDMA technology.
   The Qualcomm executives had read Cyberpunk, the first third
of which detailed Mitnick's exploits up through his arrest in
1988, and wanted to know if Markoff had any information that
would help them confirm what they believed-that Mitnick was
behind a recent, well-executed computer break-in during which
someone had stolen copies of the software that controlled
Qualcomm's cellular telephones.
   The theft had begun with a series of phone calls to new engi-
neering employees from someone who claimed to be a
Qualcomm engineer from another group. He was traveling, he
said, and needed access to a particular server but had forgotten
his passwords. Trying to be helpful, the new workers were only
too happy to oblige. With passwords in hand, it was only a mat-
ter of the caller's logging onto Qualcomm computers, which were
connected to the Internet, and downloading the source code for
the new phones. When they discovered that their security had
been compromised, the Qualcomm executives notified the FBI
and were eventually put in touch with the Bureau's Los Angeles
office. A group of its agents was already on a case involving the
theft of cellular telephone software from more than a half-dozen
companies, including Motorola and Nokia. The FBI agents
showed up at Qualcomm, took down the evidence in their note-
books, and left. Weeks went by, and nothing happened. The
Qualcomm executives called repeatedly asking if any progress was
being made in the case, but they found the FBI was unwilling to
tell them anything about the investigation or their suspect. All
they would say was, "Read Cyberpunk."
   The increasinglyfrustrated Qualcomm group tried to figure out
more about their break-in on their own. How had the caller been
able to systematically identify new employees, who would be
the most likely to unwittingly give away company secrets? They
finally concluded that someone had snuck into their building
244                          PURSUIT
and taken a copy of their monthly in-house newsletter, which
routinely contained names, photos, and short biographies of new
employees. Qualcomm had long been dominated by an engi-
neer's culture, based on mutual trust and a shared team spirit, but
the theft made the executives feel that the company was under
siege, and created an atmosphere of paranoia within it. At one
point, perhaps hoping to find an employee to "chat" with, some-
one called sequentially through an engineering work area, and
several anxious Qualcomm engineers stood and listened as the
phones rang one after another.
   It wasn't clear to the Qualcomm engineers what the thief
intended to do with the software. Simply possessing it, even the
source code, would not permit someone who wanted to tamper
with the new digital cellular network to get free calls or clone
existing phone numbers, as would have been possible with older,
analog technology. Conceivably, the Qualcomm executives told
Markoff, it would be possible to sell the software, perhaps in Asia,
to some black-market counterfeiter who wanted to make cheap
knock-offs of their phone, but this seemed hardly worth the
effort. Still, according to the FBI, someone was going to a great
deal of trouble to steal software from all of the major cellular-tele-
phone makers. Why, they wondered.
   Except for the brief 1991 incident, none of this affected me
directly until October 1994, when Mark Lottar had some of his
Oki cell phone software siphoned from his computer. He warned
me to be on guard, and sure enough, several days later Andrew
began seeing probes on Ariel. At one point someone started elec-
tronically scouting the network portals to our systems. Andrew
could see that the interloper was testing widely known network
security loopholes in an effort to enter our machines. To repel the
invader, he began closing off various potential access routes in
response to the attacks. The probes continued until nearly mid-
night one night. Somebody was clearly interested in our com-
puters, and because of what Mark had told us we later guessed it
might be Kevin Mitnick. Piece by piece the puzzle had slowly
fallen into place. From the transcripts we had seen at Netcom, I
could see Mitnick was aware of me and although he didn't know
it yet, now I was on his trail.
I was awakened by the beeping of my pager sometime on Saturday
morning. In the darkened hotel room I reached over, and peering
at its display I saw it was John Markoff's home number.
   "What happened last night?" he asked after I groped for the
phone and returned his page.
   "We got a trace. We have a phone number. I think you should
come down and see some of this. We have transcripts of a con-
versation he was having with someone in Israel and they were
talking about you. I want you to look at them."
   "Where is he?"
   "The evidence suggests that he's in Raleigh, North Carolina."
   "Where are you.;>"
   It was a good question. I told him I was in a Residence Inn
somewhere in San Jose. I switched on the light and read him the
hotel's address. He said he would drive down and meet me in an
   I hung up and rolled over and went back to sleep. Forty-five
minutes later I got out of bed and stood in the shower trying to
wake up and preparing my strategy for the day. A phone num-
ber was a good clue, and it would give us something to follow.
But it was also no more than that-a single clue. Knowing our
adversary was Kevin Mitnick, I realized any given phone num-
ber in itself was likely to be of limited value. I suspected that
Mitnick may have been trying to mask his location electroni-
cally by tampering with telephone company switching equip-
246                         PURSUIT
ment, so that tracing attempts would give false information. In
1988, when FBI and state law enforcement officials had tried to
track him down in California, their own telephone tracing
efforts had led them badly astray. One number supposedly
belonging to Mitnick ended in a raid on a Southern California
apartment where police and telephone security investigators
found an immigrant cook watching television in his under-
   In the movies you get a phone trace and from it an address and
then you're done. But in real life tracing through a telephone net-
work is a much more subtle and less predictable process. Placing
a call is similar to giving someone directions to find a given
address: go down this street three blocks, and then turn right, and
so on. Tracing, in contrast, is like following directions backwards,
and it can be a frustrating exercise. As I stood in the shower, I
knew that I couldn't be certain Kevin was in Raleigh: the trace
could be wrong, or the phone call might be simply passing
through the cellular phone company switch en route from some-
where else.
   Mitnick's last arrest, in 1988, came only after his partner
DiCicco confessed to a DEC investigator. I'd heard people in the
computer underground say that the lesson he claimed he'd
learned from that incident was that in the future he would oper-
ate solo, minimizing the chance of betrayal.
   From the past week's surveillance I could tell he was still pret-
ty cocky, a bit sloppy, and a creature of habit. And from what I'd
seen so far he didn't seem to be as brilliant a hacker as legend
  Without knowing it he had made the same mistake as Mr.
Slippery, the protagonist of Vernor Vinge's wonderful 1987
cyberspace classic, True Names: he'd accidentally revealed his
identity. In his novel Vinge describes a virtual world of powerful
computers and fast networks much like the one through which I
was chasing Mitnick. And the first rule of that world was to keep
your True Name in the physical world a secret.
  Although he was making an effort to remain elusive by enter-
ing Netcom's network from different cities, he had grown lazy,
and his repeated use of the Raleigh POP was an indication that
he was coming to believe he could operate with impunity. Of
                    "TACT1CAL NUCLEAR RANGE"                     247
    course, I realized that I might be the one who was too cocky. It
    was possible that he had done something sufficiently arcane to
    protect himself so that he didn'thave to worry. My hunch was that
    he was betting that the cellular telephone companies were more
    concerned about toll fraud-stealing long-distance time-than
    about making fraudulent local cellularcalls. He was gambling that
    if he laid low and made only infrequent long-distance calls, he
    would avoid their attention.
       I sat on the bed in front of my RadioMail terminal and read
    my e-mail from the day before. One message immediately
    caught my eye-another request from David Bank, the San Jose
    Mercury News reporter. In recent days I'd received numerous
    pages from him, which I'd ignored. He clearly wouldn't give up
    on the story.

    From: Dbank@aol.com
    Received: by mail02.mail.aol.com
    ( id AA22563; Fri, 10 Feb 1995 21:35:42 -
    Date: Fri, 10 Feb 1995 21:35:42 -0500
    Message-Id: {950210213540_18424375@ao1.com}
    To: tsutomu@arie1.sdsc.edu
    Subject: SJ Merc News questions
    Status: RO

           Greetings. I'm sorry we didn't get to connect on
         Thursday or Friday. I'm still interested in meeting
         with you in person and am able to come to San Diego
         if that's easier.
           The gist of the story is that there are a number of
         people who had clear motives for breaking into your
         computer. As it happens, one of them is you. I don't
         mean any disrespect but we need to talk.
           Please call me at home on Saturday or leave a mes-
         sage at work so we can make an arrangement.


      Well, he might need to talk to me, but I didn't need to talk to
    him. The idea I'd broken-in to my own computers and then
248                         PURSUIT
detected the break-in in order to attract attention was off the
deep end. I would have to be certifiable to have presented a tech-
nical paper at a conference sponsored by the NSA less than three
weeks later.
   In any case if he wanted to write his story and jump off that
particular cliff! was quite prepared to let him. I had no intention
of calling him back or meeting with him anytime soon, no mat-
ter how aggressive he was in pursuing me.
   I was still drying my hair when Markoff arrived. While I was
getting my things together I described what had happened on
Thursday and Friday. We talked about the Israeli connection
with the student, jsz.
    "I think the Israeli connection is significant," he said. "If I
were a foreign intelligence agency or anyone for that matter who
wanted to steal technology from United States companies, what
better cover than to have a fugitive computer felon as a cut-out?"
   Markoff was sitting on my bed fiddling with the RadioMail
terminal. A single light was on in the tiny convenience kitchen,
but the room was still quite dark. Although it was gray outside I
hadn't bothered to open the curtains.
   "Maybe this is actually a Mossad operation," he continued.
"Say this guy befriended Kevin on Internet Relay Chat, or
through Hacktic in the Netherlands. Now he's goading him into
attacking various American computers. Mterwards they share the
   I didn't particularly see it. It would be easy for jsz to disguise
his identity in the Internet, and he could easily be connecting to
the school's computers from anywhere in the world. And in any
case, why would an Israeli intelligence agency have such a great
interest in cellular telephone software and development tools? It
seemed more plausible to me that Mitnick believed that by hack-
ing the cellular telephone code he could in effect make himself
invisible, for he had a huge stake in not being captured. Another
possibility may have been that he was involved in industrial espi-
onage in some way-perhaps he was stealing the software for
someone who had a real use for it.
   It was 1 P.M. by the time we left the hotel. Robert wasn't due
back at Netcom until later in the day, and we had arranged to
meet Mark Seiden for lunch somewhere between his San Mateo
                     "TACTICAL NUCLEAR RANGE"
    home and San Jose. We settled on Buck's, an informal Woodside

    restaurant and bar frequented by the venture capitalists and
    Silicon Valley CEOs who live in the exclusive bedroom commu-
       While we waited for Seiden at Buck's, I paged Kent Walker.
    When he returned my call, I said I would find a landline and ring
    him back. I walked across the street to a phone booth, told him
    about the Israeli connection, and gave him a quick summary of
    where we stood in our efforts to get a trace. We agreed to meet in
    Menlo Park at Seiden's office, since Walker was on his way to
    Stanford for a meeting and could come by and talk to us after-
    wards. He was planning on leaving the Justice Department in
    only three weeks, and I realized that he was hoping to see this case
    solved before he left government service.
       I next called Levord Burns, who had been in contact with
    Sprint Cellular, one of the two cellular phone providers in
    Raleigh. The GTE technicians had told him the call had come
    from Sprint. He'd spoken to a technician at Sprint, who had told
    him the phone number didn't belong to Sprint, that it actually
    belonged to GTE.
       "The number's funny," he said. "It doesn't go anywhere."
        This didn't make any sense to me, because a telephone num-
    ber has to go somewhere. My first thought was, Who's being inept
    here? "Is the call from Sprint or isn't it?" I asked him impatiently.
       I listened as Levord tried to repeat what he had heard from the
    Sprint technician.
       "I'm sorry, but I don't think you completely understood what
    he told you." I said, as politely as I could. "I want to talk to the
    people at Sprint directly."
       He said he would prefer to pass my message back to the cellu-
    lar technician himself
       "Lcvord, that isn't going to work," I responded. ''I'm sorry, but
    I need to talk to him directly."
       Although initially he balked at giving me their phone number,
    after further cajoling he agreed to try to find the Sprint engineer
     and set up a conference call for the three of us.
       Seiden showed up and we all sat down in a booth and ordered
     lunch. He told us about his own run-in with the Colorado
     SuperNet administrators. While he was monitoring, Mark was
250                        PURSUIT
also able to see attacks against CSN, because some were being
launched through Internex. He had called and wound up
speaking to a different person than I had spoken with, and
Mark warned him that an intruder was messing around with
the CSN computers. He described how he'd been able to see
their operating system kernel modified and then watched while
the intruder restarted the computer. Mark had several sugges-
tions for them and several questions as well, but the technical
support staffer at CSN was not about to accept this story blind-
ly, and told Mark, "I'd like your middle initial, your date of
birth, and your social security number."
   "What!" said Mark, "Why do you want that kind of informa-
   "I want to do a full NCIC scan on you before I call you back,"
came the response. NCIC is the National Crime Information
Center database which is supposed to be accessible only to law
enforcement people.
   "NCIC?" Mark was stunned. "Why would you be able to use
   "I have my contacts," the technical support person responded.
   Obviously, his contacts were the L.A. FBI agents who believed
they were closing in on Kevin Mitnick in Colorado.
   "I couldn't believe what I was hearing," Mark said, but he gave
the staffer the information and then hung up. Hours later, when
there was still no reply from CSN, he called him again.
   "What the hell is going on here anyway?" he asked. "Don't you
realize this guy just rebooted your machine?"
   It was futile. As Andrew and I had already concluded, Mark
decided he was wasting his time dealing with the people at CSN.
   Mark had spent some time carefully examining the files of
stolen material that had been stashed at the Well. After watch-
ing the intruder attack Internex repeatedly, each time bringing
over a range of system-cracking tools from the dono account on
the Well, Mark had decided to go into the account himself, and
download the entire directory so that he would be prepared for
whatever tools he was going to be attacked with. In one of the
stashes he found sniffing sessions from CSN that indicated
their administrative computer had been sniffed; as the files con-
tained both user and administrator passwords. One of the other
                "TACTICAL NUCLEAR RANGE"                   25I
things he learned, which was an obvious privacy issue, was that
Colorado SuperNet kept their customers' social security num-
bers. If you have somebody's name, address, telephone num- ,/
ber, social security number, and a credit card number you have
everything you need to make a royal mess out of his life.
  While we were waiting for lunch I paged Kathleen
Cunningham, the United States marshal, and a short while later
she called me back. This time I walked to the pay phone at the
back of the restaurant. I needed more information on Mitnick's
m.o., and I was hoping the marshal would be more forthcoming
than the paranoid FBI agents we'd been dealing with.
   I was in luck, for Cunningham was quite willing to give me
information about her efforts to capture Mitnick on an out-
standing federal probation violation dating back to late 1992.
She told me the FBI had sent a surveillance team with a
Triggerfish radio direction-finding unit to Colorado to track him.
   "Kevin is misguided, but he's not particularly dangerous," she
   She seemed to feel sorry for the fugitive and to regard him as
a poor screwed-up kid, whom it was her job to find. She sus-
pected that he was still in contact with his family, and she said
she had talked to them recently in an effort to get a plea to
Mitnick to turn himself in. We talked about the time in Seattle,
where he'd narrowly escaped local police and telephone securi-
ty investigators, largely because they hadn't known who they
were watching. She had learned that a McCaw Cellular investi-
gator and a telephone company security consultant had tracked
Mitnick for several weeks as recently as last October. They had
followed him on foot as he walked around his neighborhood
carrying a hand-held cellular phone and athletic bag, and
trailed him into a Safeway store and to the local Taco Bell. On
several nights they'd gone as far as walking up to his apartment
door (the name on the mailbox was Brian Merrill) and listened
to his phone conversations where they overheard him talking
about cracking passwords.
   On another occasion they listened in on his cellular phone con-
versations and heard portions of a conversation about getting
even with somebody's manager.
   "We'll really fuck them up," Kevin told his friend.
252                         PURSUIT
  He also mentioned Denver, indicating that he had been there
  Mter Mitnick fled, the police made an inventory of what they
found in his apartment. The evidence they collected included
gear for making illegal clones of cellular telephones. They also
found a portable computer, as well as a $1,600 medical bill for
treatment of a stomach ulcer and a prescription for Zantac, an
ulcer drug. On his kitchen table they found a radio scanner and
Aerosmith and Red Hot Chili Pepper CDs.
   Cunningham said that the FBI apparently believed that
Mitnick had recently been in San Francisco, at least briefly. An
FBI agent had been listening to a telephone conversation of an
associate of Mitnick who lived in the BayArea, in which the man
had turned away from the phone to talk to somebody in the room,
whom the agent had clearly heard him address as, "Hey Kevin."
  After I had been on the phone for about twenty minutes
Markoff came back and told me my soup was getting cold. I
thanked Cunningham for her help, and we agreed to stay in touch
as things developed.
  When I came back to the table Mark described another
intriguing clue: the Paul Kocher connection. Mark had looked up
Kocher after finding mail from between February and March of
1994 taken from his computer. Paul Kocher, a senior biology
major at Stanford University, had been interested in code-break-
ing since junior high school. He had become an ace cryptogra-
pher as a hobby and then turned it into a business on the side.
He was consulting for both RSA Data Security, Inc., a Silicon
Valley company that is a dominant force in public key cryptog-
raphy, and Microsoft.
   He'd also written a paper with the Israeli cryptographer Eli
Biham sketching out a way to break PK Zip, a widely used
compression and archiving software program that has a built-
in encryption feature. Biham is in the computer science
department at the Technion, a prestigious science and engi-
neering school in Israel, and is known as one of the world's best
cryptographers. In December 1991 he had also published a
paper with Adi Shamir, another Israeli cryptographer, that
had laid out one of the first partially successful research efforts
demonstrating potential weaknesses in the U.S. Data
Encryption Standard, the nation's coding standard used by the
                 "TACTICAL NUCLEAR RANGE"                      253
government, industry, and by banks, and other financial insti-
  After Kocher and Biham had published their report, Kocher
had also posted on the net a portion of it describing the method
for breaking passwords beginning with the letter z. His intention
had been to prove that the KocherlBiham technique was a suc-
cessful way to break the code, without making it available for all
passwords. It appeared that Mitnick had seen the posted materi-
al and had decided to target Kocher's files in order to get the com-
plete version of the program.
   Mark called Kocher, and the Stanford student drove up to his
home in Belmont to look at the files. He had a remarkable story
to tell. About the same time that my computers had been
attacked in December, Paul Kocher had gotten an electronic mail
message from Eli Biham, "Paul, can you send me a copy of the
PK Zip decryption program? I really need a copy of it for my
   Kocher didn't respond to it, because the request seemed out of
character. Biham surely knew it was a violation of United States
export control laws to transmit cryptographic software out of the
country without an export license.
   A week later a more strident note arrived from Biham saying,
"Paul, where is that source code I asked for?"
   This time Kocher responded with a note back to Biham saying,
"Eli, you know about the cryptography laws much better than I
do. Why are you asking me for this?"
   A few days later he received a reply from Biham addressed to a
long list of people saying, "Anyone who received mail from me
during the last month should distrust it. I have reason to believe
my account was broken-in to and taken over."
   Just as we were getting up to leave, Laura Sardina, one of Sun
Microsystems' first employees and a longtime friend, walked by.
She is a person who really knows how to get things done in the
company, and I asked her if! could borrow some SPARCstations,
thinking that if this turned out to be a protracted hunt, we were
going to need more hardware to set up monitoring in different
locations. She wanted to help and told me to come by her office
 on Monday.
   After we left Buck's we followed Seiden to Menlo Park to meet
with Kent Walker, driving down Woodside Road, which stretches
254                        PURSUIT
down to the Bay from the hills, and where the Valley's most
successful computer moguls have their mansions and horse
   Dressed in jeans, Walker looked even younger than he did in
his weekday business suit attire. I recounted what we'd learned in
the past two nights and continued to press him for more help in
getting warrants from the phone companies in Denver, as well as
a trace order with Sprint Cellular in Raleigh.
   "I can't help you in Denver," he said, "but if you want a trace
in Raleigh you've got it."
   It was now past five in the afternoon and was already starting
to get dark. At Netcom, Robert and Andrew had resumed their
watch, and we got back on the freeway and drove down to San
Jose to join them. When we arrived I found that we had what
might become a more pressing problem. Andrew had called Pei
at the Well, who told him that, on the previous night at about
ten-thirty, their monitoring had revealed the password Mitnick
was using to log-in to his account at escape.com. After he had
left, Pei had decided on her own that she would log-in as Mitnick
and look around for herself.
   The problem was that in doing so, she might have given away
our element of surprise. Most computer operating systems alert
the user each time he connects as to the exact time he logged-on
previously. It's a simple security precaution that can tip a com-
puter user off if someone else is using his account.
   "Why did Pei do this?" I was irate. "What was she hoping to
   "I have no idea," Andrew said.
   "Did she clean up after herself?" I asked.
   "No," he replied.
   I couldn't believe that anyone would do anything so foolish,
particularly someone who supposedly understood computers and
computer security issues. Our problem now was that unless
Mitnick was totally careless, he would discover that someone was
aware of his presence the very next time he used his account on
   Even worse, if we were unlucky and he was running his own
sniffers on escape.com or the Well, he would know exactly who
was following him. There was nothing we could do to correct
                "TACTICAL NUCLEAR RANGE"                      25 5
Pei's mistake, and our only option was to watch and wait. Maybe
we'd be lucky.
  "Call her back and explain to her what she did wrong," I said
to Andrew. ''Ask her to please give us just a few more days before
they go waving a red flag in Kevin Mitnick's face,"
  At our portable computer set up outside of Robert's office, I
played back for Markoff the keyboard conversation between
Mitnick and jsz from Friday, and when he saw that the fugitive
thought it would be possible to falsify a New YOrk Times story by
breaking in to nytimes.com he laughed. "If they only knew," he
said. "Times management is so paranoid about something like
this happening, that the Atex editorial system has no interactive
connection to the Net."
  From reading the conversation between Martin and jsz again
we gained another clue. Martin had mentioned seeing the movie
Sneakers, and Markoff recognized the significance of the marty
user name and control-fbishop. Kevin Mitnick, it seemed, had a
continuing obsession with the actor Robert Redford. First there
was Condor, and now it appeared that Mitnick had adopted
another one of Redford's guises. In Sneakers Redford had played
Marty Bryce, an antiwar activist and a computer hacker who'd
gone on the run in the 1960s and years later had taken the name
Marty Bishop. In the film, Bishop has created his own computer
hacker's "dirty dozen," a group that ends up doing contract work
for the National Security Agency.
  The Marty connection was one more confirmation that it was
Mitnick we were tracking, and I hoped that it might also be an
important clue about his location. I called a friend in Boulder
and asked him to check the television schedule to see if the movie
had shown recently, for it might indicate what region Mitnick
was in. Unfortunately, it turned out that Sneakers had played on
network television throughout the country.
   The log-in records indicated that Mitnick had last been on
Nercom in the middle of the afternoon. We looked back
through our filter data and found that he had connected to a
computer called mdc.org, the Internet domain for Lexis-Nexis,
the online database company. He used a stolen password to
access their current news database and then typed the follow-
ing search command: MITNICK w/30 KEVIN. He was looking for
25 6                         PURSUIT
any occurrences of his name in recent news stories! Our tran-
script showed that he had looked at the complete text of one
story after scanning the headlines of the most recent additions
to the database.

             LEVEL 1 - 46 STORIES
             1. Newsweek, February 6, 1995 , UNITED
       STATES EDITION, BUSINESS; Pg. 38, 270 words,

               2. Deutsche Presse-Agentur, January 24, 1995,
       Tuesday, International News, 614words, U.S. hunts mas-
       ter computer "cracker", Washington

               3. United Press International, January 24, 1995,
       Tuesday, BC cycle, Washington News, California, 605
       words, U.S. hunts master computer 'cracker', BY

               4. United Press International, January 24, 1995,
       Tuesday, BC cycle, WashingtonNews, 606 words, U.S.
       hunts master computer 'cracker', BY MICHAEL KIRK-
       LAND, WASHINGTON, Jan. 24

               5. U.S. News & World Report, January 23,
       118, No.3; Pg. 54,3666 words, Policing cyberspace, By
       Vic Sussman

               6. Pittsburgh Post-Gazette, December 20, 1994,
       Tuesday, SOONER EDITION, Pg. B1380 words, Six
       inmates sue, charging jail beatings, Marylynne Pirz, Post-

               LEVEL 1 - 2 OF 46 STORIES
               Copyright 1995 Deutsche Presse-Agentur
               Deutsche Presse-Agentur

               January 24, 1995, Tuesday, BC Cycle
               23:04 Central European Time

               SECTION: International News
                 "TACTICAL NUCLEAR RANGE"                              257
               LENGTH: 614 words
               HEADLINE: U.S. hunts master computer

               DATELINE: Washington
                 U.S. law enforcement authorities asked the
       public for help Tuesday in their effort to track down a
       legendary master manipulator of the information super-

                Officials said Kevin David Mitnick, 31, origi-
       nally of Sepulveda, California, is using his hacker skills to
       stay one step ahead of the law ...

   By about 7 P.M. everyone was hungry. Robert, who had been
so enthusiastic about the chance to track his nemesis on
Thursday, was now sleepy and taciturn. It looked as if we were
facing another long night, so Markoff and I decided we would
go out and find dinner for the four of us. We drove for several
blocks past movie theaters and shopping centers until we finally
located a Round Table pizza parlor. We ordered two pizzas, and
while we waited we took seats at a long table in the mostly
vacant dining room. We talked about Julia's weekend trip and I
told him I'd felt a sense of relief when Julia had left Friday
evening, but I also missed her.
   Levord finally called back shortly after we returned to Netcom
and said he would set up a conference call in a couple of minutes.
When he rang back I could barely hear the Sprint technician on
the other end of our call.
   "Tsutomu, this is Jim Murphy, I'm a communications engineer
with Sprint Cellular in Raleigh."
   His voice was faint because it was a conference call and I asked
if he was using a cellular phone. He was.
   "Excuse me, but I really don't want to have this conversation
while you're on a cell phone," I said. Levord had set up the call,
but I was amazed that he hadn't considered the potential security
problem. A radio scanner had been found in Mitnick's apartment
in Seattle; hadn't anyone realized that he might easily be able to
intercept this conversation?
   Murphy explained that he was out in the field and that it would
25 8                        PURSUIT
take him about ten minutes to drive back to the cellular company's
main switching office. When we resumed the conversation
Murphy was still so faintly audible on the other end of the line
that we were shouting back and forth at each other. Neither of us
had a good explanation of why both the Sprint and GTE switch-
es were showing the call should have come from the other one,
but we both realized it wasn't possible. I explained to him whom
we thought we were dealing with, and that Kevin Mitnick had a
fifteen-year history of tampering with telephone company
switches. He was incensed at the idea of someone messing with
his switch, and as we talked it turned out that Murph, as he pre-
ferred to be called, was in fact very sharp, so that we immediate-
ly dropped into technical detail.
   I began by asking him questions about the telephone switch
the Sprint system was using. Telephone company switches are
actually just computers with their own specialized operating sys-
tems. Often they have dial-up POftS for remote diagnostics and
maintenance. Frequently phone phreaks and members of the
computer underground have used these ports as back doors to
tamper with the switches. They can get free phone calls or create
chat lines which anyone can dial in to. The Sprint machine was
a Motorola EMX 2500, in tandem with a DSC 630 switch, a
device about which I knew nothing. I've had some experience
with small telephone company and PBX switches, but not much
with large central office switches like this. Murph gave me a tuto-
rial on how his switch worked and what kind of data he had avail-
able. He had to be careful, because while we had a warrant for
GTE information, Kent had not yet prepared one for Sprint, so
Murph was limited in what kind of actual caller data he could
offer me.
   I asked him about the GTE number. It turned out the number
that had been captured by the GTE trap-and-trace warrant was
919-555-2774. "Is this a cellular number, or is something in the
Originating Number Identification information getting scram-
bled?" ONI is also used to provide Caller ID, the feature that
passes the caller's telephone number over the network to the
called telephone and identifies the caller.
   "It's not one of our numbers," he answered. "That prefix isn't
even a cellular phone prefix."
                     "TACTICAL NUCLEAR RANGE"
       At this point I knew something was amiss. Normally, techni-
    cians can get call-tracing information by looking a number up in
    a database that is maintained at the telephone switching center.

    If the number is a local number controlled by the switch, the
    database will show the exact set of phone wires the call is com-
    mg m on.
       In this case the GTE calling records were showing the call came
    from a permanent digital T-l connection between the GTE
    switch and the Sprint cellular switch across town, used for rout-
    ing calls between the two switches. A call coming in to a switch
    from another switch comes in on what is referred to as a trunk
    line-in this case the T-l. It can carry twenty-four telephone calls
    simultaneously. The switch looks in its database of translation
    tables and routes each individual call according to the informa-
    tion it finds. If it is a call being delivered locally, the translation
    tables will direct the call either to a particular telephone line, or
    in the cellular world to the equivalent, known as a Mobile
    Identification Number (MIN).
       As we talked, Murph kept checking his switch to see ifhe could
    find anything obviously amiss or something that had been tam-
    pered with. While I waited on the other end of the line he
    explored the innards of the computer, examining its translation
    tables while he gave me a running commentary of what he was
    looking at. He said he had a theory that Mitnick might somehow
    have created a special number that would route his calls through
    the cellular switch, and then on to a Netcom local dial-up num-
    ber. Every phone number has a direct route as well as an alternate
    route, and he wondered if one of the alternates had been messed
    with. He spent a long time probing his database to see ifhe could
    find any evidence of such a hidden route.
       Nothing obvious showed up, however, and we began looking
    for alternative explanations. Murph had records in a database that
    could be searched and sorted with many different parameters.
    However, each of these operations took up to half an hour.
       We talked about useful ways to sort through the data, and then
    it occurred to me to ask, "What happens when I dial the GTE
    trace number?" I did so and heard this eerie "click-click," "click-
    click," "click-click," which continued to repeat, getting fainter
    and fainter until it disappeared, and the call disconnected.
260                          PURSUIT
   I came back on the phone and described to Murph what I had
   "My guess is what you're hearing is the call endlessly looping
between the GTE switch and ours," he said. "Eventually the
power falls below a certain level, and the call is dropped."
   I tried it again, and this time Murph monitored it from his
switch. Again I could hear the "click-click" sound, but at the
same time I could hear the printer in his office register each time
his cellular switch tried to set up a call. "Kerchunk," "Kerchunk,"
   ''I'll be very surprised ifhe's tampered with our switch," Murph
said. "We do have remote capabilities, but all remote accesses are
logged. When Motorola, for example, connects to our switch, we
first give them a password, monitor their activities, and then
immediately change the password after the session ends."
   "Let me try something else," I said. I dialed the phone number
that was one number higher than our mysterious phone number.
On the other end of the line I heard the familiar warble of a fax
machine. Murph didn't see the call go through his switch this
time. It made me even more suspicious of GTE. It told us that
only one phone number in an entire block of phone lines had
been routed to Sprint. Something was funny about that particu-
lar phone number.
   "My guess is that the GTE switch has been hacked," I said.
   We continued to puzzle. He said he could start three simulta-
neous searches to try to find a match to the Netcom log-in infor-
mation I had, because he had three terminals.
   "Let's try a different strategy," I suggested. "How far back does
your database go and what kinds of things can you search for?"
He said he could go back as far as 3 P.M. on Thursday, February
9, and he gave me a long list of sortable categories, including call
start and end time, call duration, called number, and so on.
Looking down my list of gkremen's log-ins from the Netcom
POPs, I saw that there were several long sessions.
   "Can you search for calls of a duration of more than thirty-five
minutes on Friday?" I asked. I had decided that while it might
have been possible for Mitnick to conceal where he was calling
from, it would be much more difficult to conceal the fact that a
call was taking place at all. This was the beauty of traffic analysis.
                     "TACTICAL NUCLEAR RANGE"                      26I
    The second request I had for Murph was to search for all cellular
    telephone calls made to the range of numbers that were routed to
    Netcom Raleigh dial-in telephone numbers. Finally, I asked him
    to search for all cell phone calls to Netcom's Denver number.
       Few people use cellular modems to transmit data, so any cellu-
    lar call to a Netcom POP would be unusual. And in any case,
    given that Netcom was a local call, a long distance call to a dial-
    up number would be even more suspicious. In any case, if
    Mitnick had been making calls using the Sprint cellular system
    we should have been able to find them here, even if GTE was
    unable to trace them.
       Now I had my three questions. As he set up his computers
    Murph said it was going to take a while to do the database search,
    so I told him I would ring him back in a while and hung up.
    Shortly afterwards I realized we'd both completely forgotten that
    Levord had been listening on the line.
       Because the PBX installers were still working at Netcom and
    phones kept being taken off-line, I moved to the opposite side of
    the building from Robert's office where I settled in a vacant car-
    rel with a still-functioning telephone. I called Murph back after
    about a half an hour to check on the results of his searches.
      We started with the local calls to Netcom's Raleigh POP.
       "I think I've seen that first number," he replied.
       "Good! Can you give me all the calls to the Raleigh POP?"
       "I can't tell you the actual calling numbers because you don't
    have a warrant," he replied. "I can't give you the actual MIN-
    ESN pairs." The MIN and ESN-electronic send number-are
    the two separate numbers that define a particular cellular tele-
    phone. The MIN is the assigned cellular phone number and the
    ESN is the permanent serial number embedded in the phone.
       "I don't want the number," I explained, and told him that I was
    trying to match calls to the sessions we had seen from the
    Netcom Raleigh dial-up. I was more interested in the pattern of
    calls than in the actual data. I wasn't looking for the actual num-
    ber, I was curious to see if there was a pattern to the calls that
    Mitnick might be making to Netcom through Sprint. If we were
    lucky we might discover that all the calls came from a small num-
    ber of MINs or from the same physical location.
       We began playing a game that was a lot like the classicchildren's
262                         PURSUIT
game Battleship. He couldn't tell me what the number was, but
he could tell me if it was the same as some other number under
certain conditions.
   What I could say was, "Do you see this call at this time?" I took
two lists, the Netcom list of dial-in numbers from around the
country and the summary of gkremen's log-in sessions.
   "On Friday at 15:29 do you see a call to 404-555-7332 dura-
tion approximately 44 minutes?"
   "Yes, I have that."
   "Do you have a call of duration 49 minutes at about 20:22
your time on Friday to 612-555-6400?"
   "I have it."
   "Do they both come from the same MIN?" I asked.
   "Yes," he replied.
   "Do you have a phone call on February 11 at 02:21 to 919-
   "Yes, I have that one, too."
   I asked the same question with five more log-ins taken at ran-
dom. In each case the answer was the same: they'd been placed
from the same cellular telephone number. Occam was right.
   "So where is it?" I asked.
   Murph walked across the room to a map of Sprint's Raleigh cell
   All of the calls were coming from cell number 19, located on
the northeastern outskirts of the city, near the airport. We now
had another important piece of information: Mitnick was at a
fixed location. I thought it was unlikely that the calls would be
made while he was driving, but I had been worried that he might
be changing locations with each call.
   "Do you have sector information?" I asked. Some cellular sys-
tems can determine which direction the calling phone is actually
located in relation to the cell site -that is, the particular trans-
mitter-receiver tower in a certain area.
   "No we don't have that information, but to the east of the cell
site is Umstead State Park and to the northwest is the airport. My
guess is he is transmitting from somewhere south or west of the
cell, based on the locations of our other cells."
   It was almost one in the morning. By the time we were through
                      "TACTICAL NUCLEAR RANGE"
    we had his location narrowed down to a radius of less than a kilo-

      ''I'll fly out first thing in the morning," I told him. "I'll see you
    tomorrow. "
      He gave me his numbers and told me he would meet me at the
      Even though it was late, I called Kent back and told him it was
    more important than ever to get trace orders for both the cellular
    phone companies. After I got off the phone I remembered it had
    been hours since I'd heard from special agent Burns. It was four
    in the morning on the East Coast when I called him back to tell
    him we'd pinned Kevin down.
       "You hung up on me," he said when I woke him up.
       I suspected that he'd actually fallen asleep on us and hadn't
    noticed, but I apologized for forgetting him.
       "We have it down to within a kilometer," I told him. ''I'm fly-
    ing to Raleigh tomorrow morning, and we're going to need to get
    a radio direction finding team out there."
      It was late. All I got was a noncommittal, "Unnnh."
       I'd seen Markoff leaving a half an hour before and I called him
    on his car phone. Since I didn't know who in the Valley was lis-
    tening, I was circumspect.
       "We're within tactical nuclear range."
I got very little sleep that night. Andrew and I returned to the
Residence Inn, but I stayed up making phone calls and trying to
get things organized in Raleigh. I was trying to persuade the FBI
to send agents and a radio direction-finding team there quickly. I
wanted to go to Raleigh so I'd be in a better position to get infor-
mation and to make decisions in case our target's behavior
  At 4:30 A.M. I paged Kathleen Cunningham in an effort to get
a Triggerfish radio surveillance team sent to Raleigh. She said she
would do her best, but after I hung up I had a moment of panic
wondering whether she was going to contact the Los Angeles
FBI, and whether they were going to try to get in the way. From
what I could see, Kevin Mitnick was a creature of habit in the
network world. It was becoming apparent that he wasn't that
clever and he was prone to mistakes. At the same time it looked
as if he thought he was invulnerable. All this should have made
him relatively easy to take down. But for the FBI, who had mas-
tered traditional investigative techniques, yet was ignorant of
computers and computer networks, he might as well have been
Casper the Ghost. If the L.A. FBI did decide to take a role, how-
ever, there wasn't much I could do about it. Kent Walker was
helping me, and I'd just have to see what kind of support I could
marshal when I arrived in Raleigh.
   I called American Airlines and made a 9:20 A.M. reservation
                             RALEIGH                            26 5
through Chicago for Raleigh, scheduled to arrive there at 7 P.M.
that evening. I arranged for a first-class seat because Kent had
asked me to have easy access to an AirFone, and I wanted to be
able to stretch out and sleep.
    Feeling ill from fatigue, I got up the next morning and stag-
gered through the hotel's continental breakfast. Andrew drove me
to the airport shortly after 8 A.M., and as we pulled up to the
departure gate I asked Andrew if he would contact Julia and tell
her where I'd gone. But something else was bothering me, for I
was worried that if Mitnick had an accomplice, we were likely to
lose the stolen software and find it spread throughout the
Internet by members of the computer underground.
   "Would you put together a list of all the sites where Mitnick has
stashed software and come up with a plan for collecting evidence
and cleaning up when he's arrested?" I asked him. "Don't do any-
thing yet-let me clear the legality with Kent Walker first."
  As I settled into my seat on the plane I thought to myself, This
is weird, it feels like a movie. I'd been chasing an electronic
chimera for more than a week, and now in the last few hours it
had been transformed from some tenuous image on the Internet
to a real person in the real world. It's not the kind of situation an
academic researcher usually finds himself in. We track this guy
down, and then five hours later I'm on the first flight out to try
to locate him.
   I dozed only a little on the flight across the country. After I'd
been airborne for a couple of hours I discovered that Andrew had
managed to secretly slip some apples and bananas into my gray
daypack. It was a nice thought and it explained why my baggage
had suddenly become so heavy.
   I had a short time between planes in Chicago, during which I
called Levord again. I asked him if he'd had any luck getting in
touch with Cellular One, the other cellular carrier in Raleigh. We
would still need their help as well. He told me he still hadn't been
able to find a phone number.
   "Have you tried 1-800 CELL-ONE?" I asked him pointedly.
   "No, I haven't tried that yet," he said, clearly annoyed with the
suggestion. Things between Levord and I had begun poorly and
were rapidly deteriorating. I could tell he didn't like taking orders
from a civilian, but he was in an awkward position, since the
266                          PURSUIT
Justice Department had told him to cooperate with me. My sense
was that he was in over his head technically, and I could see that
he resented my having taken charge of the search.
   Next I called Kent and asked him about the legality of clean-
ing up after an arrest. He said that as far as he was concerned, it
was legitimate because we were protecting the victims' property.
   On our plane's final approach into Raleigh I called Murph,
who promised to meet me with one of his partners at the termi-
nal. Even though it was the middle of winter in the East, I was
still wearing California clothing: hiking shorts, a purple Gore-
Tex jacket, Birkenstock sandals, and no socks.
   Julia had told me about growing up in Durham surrounded by
the smell of tobacco, and as I walked through the terminal I was
immediately struck by its sweet smell everywhere.
    W,/re not in Kansas anymore, Toto, I thought to myself
   While I was waiting for Murph I went to a bank of telephones
and called Levord again. I still hadn't been able to extract a com-
mitment of support from him, but he was beginning to say that
he was considering showing up. I was anxious to be ready in case
Mitnick jumped from one cellular system to the other in Raleigh,
 but Levord said he was still having trouble contacting Cellular
One. I kept remembering what Kent had said about not taking
 orders from law enforcement officials, and that in our case law
 enforcement would provide legal and administrative backup for
 us. I didn't actually share that with Levord, as I didn't see any rea-
 son to rub his face in it. By the end of the conversation he agreed
 to contact a local agent in Raleigh.
    For all my persistence, I still wasn't certain Mitnick was actually
 in Raleigh. What if he had set up some kind of ingenious repeater?
 I could imagine FBI agents going to an apartment and finding
 nothing but an elaborate communications system and an alarm to
 let Mitnick know that his cutout had been discovered. It wouldn't
 send us all the way back to the beginning, but it would be one
 more door that we would have to break down.
    I was still standing at the bank of phones in the airport when
 Murph came in to meet me along with another engineer, Joe
 Orsak. Murph was a large, stocky man who looked as if he had
 played football earlier in life. He had a direct, no-nonsense man-
 ner and just a touch of a Southern accent. His companion was
                            RALEIGH                           267
even larger, a burly guy with a friendly face and a moustache.
They both seemed excited about the prospect of an adventure
that would take them away from the normal day-to-day engi-
neering and maintenance of cellular telephone switches.
Recently, they had been involved in a successful bust of an illegal
cellular ring in the Raleigh area that had been using cloned
phones and selling stolen cellular phone time for international
calls from a farmhouse outside of town, and that incident seemed
to have whet their appetites for chasing down more phone fraud.
As we went out to the Sprint Cellular truck at the curb I thought
to myself, If Mitnick believed he could hide away here in some
backwater, he clearly chose the wrong place.
   I needed to pick up a rental car, and wound up with a Green
Geo Metro in which I followed the Sprint van. Freeways are
much the same allover America, but the one thing I noticed
immediately about Raleigh was that there was a lot of road con-
struction. Everywhere I went I saw roadwork. Freeways were
being widened, or entirely new ones were being built.
   The Sprint MTSO, which Murph pronounced "mitso", using
the cellular telephone industry jargon that stands for Mobile
Telephone Switching Office, was located on the opposite side of
town, in a wooded lot on the edge of a newly developed office
park. It was dark when we arrived but I could see a concrete two-
story building behind a high chain-link security fence. Behind
the building was a tall antenna with a flashing red light on top.
   Inside we met Lathell Thomas, an FBI agent from the local
Raleigh office, who was known to everyone as L.B. He was on the
telephone when I walked in trying to get more information about
the situation into which he had been thrust. He was a black man
in his late fifties or early sixties, and he had come equipped with
the same confidential AirTel memorandum describing Kevin
Mitnick as had the FBI agents at the Well meeting. He seemed
pleasant and professional, but I could tell immediately that tele-
com fraud and computer crime were not his areas of expertise.
   I called Andrew and set up two pager codes to use in the
event of an arrest. One was a "get ready" signal. I asked the FBI
 agent for the date of Mitnick's birthday, but he was busy so I
went over and grabbed his AirTel and read the date off the
wanted wire: 080663. Next I came up with 122594, the date of
268                         PURSUIT
the first break-in to my computers in San Diego, which would
serve as the "go" signal for Andrew to start cleaning up the
stolen software.
   Andrew told me Mitnick had been on again, and they had
seen a couple of intriguing keyboard chat sessions. The first
one took place about noon with a friend who was a member of
Mitnick's old gang in Los Angeles. The friend's home and
office had been searched by the FBI at the same time Mitnick
disappeared. He had sued the Justice Department over the
search and he was publicly taunting the FBI agents who were
hunting for Mitnick. Much of their conversation was cryptic,
Andrew said. His friend referred to Mitnick as "Kremlin," and
Mitnick called him "banana." They were talking about a pre-
arranged signal that would come later and permit a direct tele-
phone conversation.
   It was an odd conversation and Andrew and I puzzled over
what appeared to be other codewords in their conversation. The
friend complained about a "mosquito," and then a while later
Mitnick typed, "hahahaha. yup I didn't understand your message
re: news, mosquito. I guess." What did mosquito refer to? Did it
mean that they were worried about being bugged? At the end
Mitnick typed, "I heard jl assistant was at hottub's." We both rec-
ognized "hottub." In messages that the same friend had posted
on Usenet conferences in the past he would include a line at the
bottom of each message referring to one of the FBI agents search-
ing for Mitnick as Kathleen "Hottub" Carson.
   After our conversation Andrew faxed me some of the material
from the monitoring sessions. That morning, Mitnick had con-
nected to escape.com with a new log-in-yoda, the Star Wars
character-and found a letter from jsz. He was warning Mitnick
his father had had a serious heart attack and that he would be
away from the net for the next three to four days.
   "One more thing," Andrew had told me before we hung up. "I
might have screwed up and sent Julia to Denver instead of
   "Oops," I responded. "What happened?"
   When Julia had left on Friday night we were still thinking the
most likely upstream location would be Denver. On Sunday
                             RALEIGH                           269
morning after I departed for the airport Andrew called her to
relay the message to come ahead and join me if she wanted to.
Since the resort where she was staying was very rustic, none of
the rooms had telephones, and the public pay phone wasn't
working, so he'd left an urgent message at the office for Julia to
page him. Andrew was worried about the stories he had heard
about Mitnick's prowess as a wiretapper and when he and Julia
finally connected they had an especially cryptic conversation so
there would be no danger of giving anything away.
   Andrew said, "Tsutomu went to the place he was planning to
go next."
   Some time after he hung up he realized that he had no idea if
she thought they were talking about Denver or Raleigh, but by
then it was too late to reach her again. There wasn't anything I
could do about it, either, for I had no idea how to get in touch
with her. I'd just have to wait and see if Julia checked in with us.
   Before long, she did page me. I called her back and she told me
that she had booked her flight to Denver and was calling me to
make contact.
   "Good," I said, "but I'm not in Denver, I'm in Raleigh." I told
her briefly what had happened in her absence and she said she
would make a reservation and be on the next flight out. A few
minutes later she called to tell me she would be on a red-eye flight
and would be arriving in the morning.
   Inside the Sprint switching center we once again began playing
a waiting game. Mitnick had disappeared from the Sprint-system.
Andrew was seeing activity at Netcom, but today Mitnick wasn't
showing up on Murph's consoles. The Sprint Motorola mini-
computer that controlled the switch was agonizingly slow at sort-
ing the call detail records we needed to match the profile we had
from the previous day's activities. After a while one thing was
clear: There was no Mitnick. The calling number from the day
before had disappeared. I suggested broadening the search net to
see if he'd simply changed his behavior or was using another
phone number. We'd wait and wait with each query. I could tell
this computer wasn't designed to do this kind of searching-it
was designed for customer billing.
   "Can you dump some of the calling data onto a floppy disk?"
270                         PURSUIT
I finally asked him. "If you can get it out of your system we can
put it on my RDI and do more sophisticated searches."
   Murph said it was possible and we started to unload his data.
But then he stopped and after thinking for a second decided to
try something else first. He called an engineer he knew at Cellular
One and asked him to look through his records for any suspi-
cious activity. We gave him a profile of things to look for, but the
technician said he didn't have any matches either.
   I'd flown all the way back to the East Coast, and now Mitnick
was starting to look a little like Houdini. If he was on Netcom,
but not going through either Sprint or Cellular One, where was
he? I was exasperated-he had to be on one system or the other.
     .try some more, "1 Sal. "H e has to be t here."
   "'T'                    id
   I found several other Netcom dial-in numbers elsewhere in the
country and read them to him. The Cellular One engineer took
another pass through his data, and a short while later he came
back to the phone and said he had caller activity matching our
description. Mitnick was on the air, but where was he?
   "I can't give you guys any more help unless I've got a warrant,"
the engineer said.
   We were blocked again by the same problem Murph and I had
on Saturday night, for we didn't have a subpoena for the Cellular
One records. Although on Sunday morning Sprint had received
a subpoena for trap-and-trace and call-record information, and a
court order permitting real time monitoring, Mitnick must in the
meantime have "tumbled" his cell phone. He'd obviously
swapped the fraudulent Sprint MIN-ESN pair for one that must
have belonged to a Cellular One subscriber. I called Kent and
with Murph's help he wrote a second subpoena, which was faxed
to Cellular One.
   However, at this point things seem to come to a halt. It had
been my plan to assemble a team of law enforcement agents, to
go to the cell site, and when Mitnick went on the air, to use hom-
 ing gear to track him down. Again the FBI had put on the brakes.
Special agent Thomas had been called out on a case he knew
 nothing about on a Sunday night and he made it clear that he
wasn't ready to make any decisions about the next step to take
without the involvement of some higher authority.
   I couldn't believe it. We had Mitnick, and could track him down
                              RALEIGH                            27I
immediately. But the longer we took to pull things together,
the more likely something was to go wrong. "This is why Kevin
Mitnick is still out there after disappearing to avoid an FBI search
in 1992," I grumbled.
    I went into the back room of the switching office and called
Kent back to express my frustration. "It's really messed up here,"
I told him. "I'm getting really tired of this."
    He was becoming familiar with my annoyance and promised
he'd make some calls to see if he could speed things up. But the
situation only became worse. After I got off the phone we began
discussing the operational details of conducting surveillance and
an arrest. Special agent Thomas assured me that law enforcement
agents would have no problem staying in contact with each other,
for they all carried scrambled radios.
    "You can't use scrambled radios," I had to explain to him.
"This guy isn't a normal criminal. He works with his scanner on."
    "He won't stick around for one moment if he hears encrypted
traffic nearby," Murph chimed in, and we eventually got our
point across.
    By now it was almost 10:30 P.M. Despite the FBI agents' hesi-
tation, we decided we could still go to the cell site and use Sprint's
diagnostic equipment to get a better fix on Mitnick's location.
Murph suggested we follow the same approach they'd taken in
their recent phone fraud investigation. Every time a cellular
phone call is set up it is allocated to its own frequency. That fre-
quency was visible to the engineers monitoring at the cellular
telephone company switch, and so they had used a system in
which each time the frequency changed, the engineers would
send it to a pager the technician in the field was carrying. The
 technician would then tune the direction-finding equipment
accordingly. It seemed like a good idea. It was unlikely Mitnick
would be monitoring both cellular and paging frequencies. Even
 if he was, it was unlikely he would attach any significance to an
 occasional three-digit page.
    We called the Cellular One technician back to help alert us
when new calls were made. He was monitoring his switch from
 home and was able to see both caller data and also sector infor-
 mation. Since there was a Cellular One cell site located immedi-
 ately next to cell 19 in the Sprint system, we were now able to
272                          PURSUIT
determine that Mitnick's calls on the Cellular One system were
being placed from a phone in the same general location as the
previous evening's calls. We were in luck!
   The phone calls were coming from the area immediately south
of the cell transmitter, confirming Murph's earlier suspicion
about Mitnick's location. Murph, Joe, and I walked over to a
large map of the Raleigh area. The transmitter was located off
Route 70, also known as Glenwood Ave. Directly to the south
was Raleigh Memorial Cemetery; to the east and southeast was
William B. Umstead State Park.
   Immediately our eyes were drawn to Duraleigh Road, which
ran almost directly south from where it intersected Glenwood.
On the east side of Duraleigh for about a kilometer stretched a
neighborhood named Duraleigh Woods. It looked like it would
be a good place to start hunting. Murph wasn't sure about
Mitnick's distance from the cell, but he drew an arc around the
antenna site and said he would probably be within that range.
   On the floor of the backseat of Joe Orsak's van was a device
about the size of a desktop PC called a Cellscope 2000, which
was actually an amateur radio transceiver connected to a note-
book personal computer. Used by cellular phone companies to
test signal quality, it could function as a radio direction finder as
well. Orsak also had a hand-held Yagi antenna connected to the
Cellscope, that he could hold inside the cab of his van. The Yagi
wasn't designed for radio direction finding, but it would work in
an approximate fashion.
   Mark Lotter's software was running in my HP 100 palm-top
PC in combination with a hand-held Oki 1150 cellular tele-
phone, a setup that performed some of the same functions, but
in a smaller and cheaper package. It wasn't directional, but for
my application it didn't matter. In the cellular telephone world
the base station to cellular telephone link is called the forward
channel and the link going back from the cell phone to the base
station is called the reverse channel. The Cellscope could mon-
itor either channel, but not both simultaneously. Using it in
tandem with my hand-held system, however, we were able to
track both sides of a call.
   Mark and I had designed a custom cable to connect the com-
puter to the Oki phone. Inside it was a microprocessor chip that
did the data conversion between the Oki and the HP computer
                            RALEIGH                           273
so that the two devices would be able to talk to each other. The
little chip has as much processing power as the first personal com-
puters. It's emblematic of a story Danny Hillis likes to tell.
During the 1970s at a computer conference at the New York
Hilton a speaker made a wild estimate about the number of com-
puters that would be in the world a decade later. Somebody in the
audience stood up and said, "That's crazy! For that to be true.
there would have to be a computer in every door!"
   A decade later Hillis returned to the Hilton for another con-
ference, and sure enough, there was a computer in every door-
in the Hilton's newly installed electronic door locks!
   As we drove to the cell site I began to assemble my gear and to
tinker with the Cellscope while Joe gave me instructions about
how to use it. Intercepting cellular phone calls with a device like
this is outlawed for private individuals by the 1988 Electronic
Communications Privacy Act, but cellular phone companies have
an exemption to monitor in order to detect and prevent fraud.
   Anticipating the possibility of a long stake-out, we pulled into
a 7-11 and I grabbed some food and something to drink while
Joe got a cup of coffee. The Cellular One engineer reported that
Mitnick wasn't active, so we drove to the cell site and waited.
Special agent Thomas had followed us in a conspicuous FBI
Crown Victoria sedan. We parked in front of a windowless room-
sized concrete block house hidden behind another chain-link
fence. Inside were banks of cellular radio transceivers to handle
the phone calls through the cell.
   Joe and I took the van out for a short drive to test the moni-
toring gear and get the lay of the land. We asked special agent
Thomas to wait until we returned, but when we came back about
twenty minutes later the Crown Victoria was gone.
   Around 11:30 P.M. Markoff paged me. I'd called him from the
airport in San Jose before I left and he'd flown to Raleigh sev-
eral hours after me and checked in at the Sheraton Imperial
hotel near the airport. I put Joe on the phone and he gave
Markoff directions to the cell site. While we waited for him to
show up we went back outside and we both turned on our scan-
ning gear. On a Sunday around midnight in the suburbs of
Raleigh things are really pretty quiet on all of the cellular tele-
phone frequencies.
    It was a cold still winter night. Joe was standing outside his
274                         PURSUIT
truck listening to the Cellscope with the Yagi antenna under his
arm and he was having trouble finding any traffic at all.
Suddenly he picked up a voice call on a Cellular One channel.
He listened for a second and immediately heard someone with a
distinct Long Island accent talking about "Phiber Optik."
           h" I"         id
  "H ey, tats It. I sal, "L's go."
  Phiber Optik was the cracker who had served a year in prison,
and who was now working for Echo, a New York City on-line
service as a system administrator.
  We both jumped into the van and raced out the driveway to
the cell site and turned onto the street. A car was coming slowly
down the street toward us.
   "I'll bet that's John Markof£" I said.
  Joe flashed the high beams of his van, and as the car pulled up
next to us, I recognized Markoff behind the wheel.
   "Park your car and get in, we got him on right now!" I yelled
out the window. He pulled into a parking lot and jumped into
the back seat. The voice with the Long Island accent was coming
out of the speaker of the Cellscope unit. We could only hear one
side of the conversation, the side that was coming from the cel-
lular base station; the cell phone itself was too far away and too
faint to hear.
   "I recognize that voice!" Markoff said immediately. "That's Eric
   I'd heard of him. As the editor of 2600 he had frequently
defended Kevin Mitnick in public, arguing that Mitnick was sim-
ply a misunderstood and mistreated computer hacker who broke
into systems out of curiosity. There were no victims when a hack-
er stole software, he claimed. We could hear Corley chatting with
someone about how to improve his public image.
   Several years earlier, 2600 h~d published Mitnick's own rebut-
tal to Cyberpunk, the book Markoff had coauthored. Mitnick
claimed he'd been framed by his partner Lenny DiCicco. Now as
we listened to the side of the conversation being broadcast by the
Cellular One transmitter, it sounded as if Corley was counseling
the person on the other side of the phone on how to deal with
being persecuted by law enforcement agencies. I wondered if
Corley knew that Mitnick was still lying to people, reading their
e-mail, and stealing software.
                            RALEIGH                           275
  Joe steered the van out onto Glenwood Ave. and then turned
right and headed south on Duraleigh Road. As he drove the
Cellscope locked onto the reverse channel going from the cell
phone back to the cell-site antenna and we briefly heard the voice
on the other side of the conversation. Although Markoff had spo-
ken with Mitnick years ealier on the phone and once heard him
speak as a computer security "consultant," he couldn't identify
the second voice definitively as Mitnick's.
   I watched the signal strength display, until it suddenly dipped.
"You've passed it," I said.
  We continued picking up snatches of the voice until at one
point we heard him saying "bye" to Corley and asking if he'd still
be up at 5 A.M. The call then disappeared into static.
   I thought to myself, Lets bepatient. I was confident now that
Mitnick was actually nearby. He might have been able to set up
an elaborate cutout scheme with a pair of data modems, but it
would have been much harder to build a relay that would handle
voice and data. Joe looked for a place to turn around, and we
cruised slowly back up Duraleigh waiting to see if we could pick
up another call.
   As we neared the intersection, we could see a succession of rea-
sonably new low-slung apartment complexes. On our right was a
shopping center and gas station. We looked over Joe's engineer-
ing maps as we drove. It seemed likely that the signal was com-
ing from somewhere inside one of the apartments. We picked the
farthest one down the road and drove into its parking lot while
we monitored. It was easy because there were no other conversa-
tions in the cell. It was almost 1 A.M.
   Our monitors now captured another call while it was being set
up. This time we could hear the whistle of a modem, which
meant it was a data call. On my display I saw the MIN, the cel-
lular telephone number, 919-555-6523. I quickly programmed
the monitor to track it in the future.
   The signal was strong. Somewhere within a few hundred meters
of where we were driving, Kevin Mitnick was sitting, probably
huddled over a portable computer, preparing to sniff passwords,
install back doors, and read other people's mail. Every few minutes
the signal would drop, and then after a pause of thirty seconds or
so a new call would start.
27 6                        PURSUIT
    "The poor bastard," I said. "He's getting really lousy cellular
   Joe drove back onto Duraleigh turning north and immediately
 turned into the driveway of a larger apartment complex called the
 Player's Club.
    As we circled we all began to feel uncomfortable. The parking
·lot was full of cars, but there were no people visible, and almost
 all of the apartments on the outside of the complex had darkened
 windows. What would someone looking out a window think if
 he saw three men in a truck circling in the parking lot at this
    We drove slowly, in a counterclockwise direction. At the rear
 of the complex we could see the apartments backed up against
 open fields. "If I was Mitnick I'd have my escape route planned
 right through those fields," Markoff said. ''I'd also set my com-
 puter up to have a good field of view out the window."
    I swung the Yagi antenna back and forth as we drove. When
 we'd entered the driveway of the Player's Club, I'd seen the signal
 strength on the Cellscope display get stronger. I could tell he was
 somewhere to out left. Now at the back of the complex, it fell off.
 The antenna didn't have much precision, because I was inside the
 truck, and I was trying to remain inconspicuous. I was also try-
 ing to keep a mental model in my mind of where the signals
 would be coming from in real space rather than car space-ignor-
 ing the orientation of the car. At the same time all three of us
 were looking for windows in which the lights were on.
    The Player's Court was circled by rows of parking spaces. The
 complex itself appeared to be a square with arms jutting out, sep-
 arated by other parking lots. As we approached the southwest
 corner of the complex the signal strength jumped again. It was
 clear that the call was coming either from along one of the arms
 or from the interior of the complex on the inside corner.
    We decided that circling the complex again would be too risky,
 so Joe drove across the street and parked in the shopping center
 lot. I was certain we'd pinned down Mitnick's location and now
 all we needed was the FBI.
    "Why don't we drive back to the cell site and see if we can per-
 suade the FBI to come out again?" I suggested.
    Back inside the concrete block house I called Murph at the
                            RALEIGH                           277
Sprint central office switch, and he called the FBI and beat on
them for a while, pleading with them to take some action. Mter
he was told they had no agents available, I called the local office
of the Bureau myself.
   "This guy is on right now," I told the FBI watch officer. "It's
like having a search light that you follow right to his door."
   "I'm sorry," he responded. "There are no agents here now. All
I can do is take a message."
   I hung up and called special agent Thomas, and got somebody
who was very unhappy to hear from me at 2:30 A.M. "I'm afraid I
can't help you tonight," he explained. "He's wanted on a U.S. mar-
shal's warrant, not an FBI warrant-it's not an FBI problem."
   I paced back and forth in the tiny room. I called Kent again
who promised me that reinforcements would be on the way soon,
but it was becoming increasingly clear that nothing was going to
happen that night. We waited for another forty-five minutes
while Mitnick's data calls came and went. Finally we decided to
give up and drive back to the Sprint switch.
   On the way back to the switch I thought about making anoth-
er phone call to Levord, but I decided a cellular phone call was
too risky. We were probably out of scanner range, but if Mitnick
was using software stolen from Mark Lottor he would have access
to the forward control channel, and he might see my telephone
number show up in the system in Raleigh. It was extremely
unlikely, but those are the kinds of things that can trip you up.
   Joe dropped me off, and I got in my car and followed Markoff
back to the Sheraton. We walked into the empty lobby at 4 A.M.
I had hoped we would arrest Mitnick tonight but now I was wor-
ried that each new delay might allow him to slip away.
Julia found me.
   She had arrived at the Sheraton at 8:30 A.M., where I had
intructed the front desk to give her a key when she showed up. I
was asleep when she slipped into my fourth-floor room, but quite
happy to be gently awakened. I was exhausted but delighted to
see her and we held each other.
   "Did you find anything last night?" she asked.
   I told her that we had almost certainly found Kevin, but as I
related the details of our surveillance, my frustration with the FBI
quickly returned. "I can't believe it," I said. "These guys are going
to let him slip through their fingers again."
   But I could see that Julia, too, was exhausted. "How was your
weekend?" I asked.
   "Surprisingly smooth," she murmured. "We communicated
more clearly than we have for a long, long time." She paused. "It
was really hard," she finally continued, "but we both agreed that
splitting up is the right thing to do."
   She slipped into bed and we were soon both sound asleep.
   When I awoke two hours later, though, my mind immediately
returned to the case, and I began making phone calls.
   The first was to Levord Burns in Washington, who told me he
was planning to make the drive down to Raleigh later that day.
 Finally, I thought, the FBI is moving into action. I asked if he was
planning to have a team of people who could stake out the apart-
ment complex.
                         THE STAKEOUT                           279
   "No, Tsutornu, it's just me," he replied, with the air of some-
one who moves at a deliberate pace regardless of what's going on
around him. ''I'm leaving in two or three hours."
   His apparent lack of concern just wasn't acceptable to me. I
didn't really regard Levord himself as a problem, so much as a
symptom of the FBI's plodding approach, so as soon as he hung
up, I decided to take my nagging to a higher level.
   Kent Walker in San Francisco once again assured me he was
still on the case and help was on the way. I could tell that he, too,
was impatient with the creeping pace of this hunt, now that we
were so close to our target. He said he would call John Bowler, an
assistant United States attorney in Raleigh, to see ifhe could look
into the case, and also promised to lean on the FBI, although we
both realized there was only so much he could do from the other
side of the country.
   Next I phoned Marty Stansell-Carum, the Justice Department
prosecutor who had been so supportive at the CMAD conference
in Sonoma, and brought her up to date. "This is exactly why I
never bothered to go to the FBI in the first place!" I concluded.
   "Who have you been dealing with?" she asked.
   "Levord Burns."
   "Oh, I know what you mean," she said. "Whenever you talk to
him, he sounds half-asleep."
   "Maybe that's because we're always waking him up in the mid-
dle of the night," I answered. I told Marty that our most pressing
need was to get a Triggerfish surveillance team on the scene so we
could figure out Mitnick's precise location. She gave me her
assurances that she would do what she could.
   Julia was up by now, only partly recovered from her night of
travel, but like me, was hungry. Around 2 P.M. we went down-
stairs and met Markoff in the Sheraton restaurant. It didn't look
like the most promising of dining spots, but because we were
staying within five kilometers of Kevin Mitnick's apartment, and
with my picture having run in newspapers and magazines all over
the country, I couldn't risk going somewhere he might see me.
And since Markoff's picture had been on the dust jacket of
 Cyberpunk, there was reason to think that Mitnick might recog-
nize him, too. From the uninspiring lunch menu, Julia selected a
white-bread club sandwich, I took my chances with grilled cheese
280                         PURSUIT
and what appeared to be vegetable soup straight from the can,
and Markoff made do with a chicken breast sandwich. He was far
more intent on getting his daily news fix than lunch, poring
through the Times and The Wall Street Journal. Julia and I were
picking at our food and chatting idly, when I got a page from
Mark Seiden at Internex,
  "What's happening?" Seiden asked, when I called him back
from a payphone in the lobby. "The clean-up's all done on this
end, but it looks like Mitnick is still loose."
  "Huh?" was all I could say.
  Seiden explained that he had been called the night before by
Andrew, who told him that, since Mitnick was about to be
hauled in, he should begin cleaning up and securing the Internex
  "This is really bad!" I exploded. "The FBI isn't close to catch-
ing Kevin yet. What if we spook him?"
  It sounded as if we already had. Seiden related that after he had
done a fairly thorough shutdown of Mitnick's back doors, Kevin
had returned through one that Mark had missed and started mak-
ing mischief, including trying to dose Seiden out of his own
account. Then, in what seemed a deliberate act of provocation,
Mitnick had deposited a 140-megabyte file calledjapboy that was a
copy of my file that Bruce Koball had stumbled upon several weeks
earlier at the Well.
  "I have no idea why Andrew told you to start cleaning up," I
said, incredulous.
  Seiden, who is a computer security pro, was angry at having
been misled into such an error. "Last time I take orders from
Andrew," he muttered. His task now, we agreed, would be to
resume monitoring Mitnick's activities on Internex for indication
of how deep his suspicions now ran. Seiden was still fuming with
indignation as we ended our call.
  I punched in Andrew's number. "What the hell's going on?"
  ''I'm sorry, I screwed up," Andrew said, knowing immediately
what I was talking about. He realized that he had misunderstood
my message from when I had called him the night before, and
that he jumped the gun with Seiden. It was just a case of being
overtired and overly optimistic.
  "Look," I said. "We're really close to catching Mitnick, but we
                         THE STAKEOUT                          28I
haven't caught him, and we may have blown the whole thing
now." I told Andrew to start monitoring Netcom for signs that
Mitnick had detected our operations there, and give me an
update later.
   Walking back to the restaurant table still shaking my head, I
told Julia and Markoff what had happened. "We don't need
screwups like this when we're so close," I said.
  Locating Mitnick, it seemed, might turn out to have been far
easier than actually catching him.

With Levord not due to arrive for several more hours, we
returned to our rooms, where Julia went back to sleep, while I
started working the phone with a new sense of urgency.
   Before long, my efforts were rewarded with a piece of good
news: Marty Stansell-Gamm told me that the FBI's Technical
Services Division in Quantico was dispatching a two-man sur-
veillance team with a Triggerfish radio direction-finding unit,
and they would be arriving in Raleigh that evening. She gave me
the SkyPager number of one of the agents, and soon I was on the
phone to him and his partner.
  As law-enforcement technical specialists tend to do, the two
agents were more interested in asking questions than answering
them. They were trying to determine what gear to bring along,
and one thing they wanted to know was whether the Sprint or
Cellular One cell sites supported NAMPS, an analog cellular
phone technology that can double the capacity of a cell site by
narrowing the frequency band that each phone uses. Cellular
companies that employ NAMPS will often compensate the cus-
tomer with lower rates for helping conserve frequency spectrum,
but using the system requires a special phone, and monitoring it
takes special surveillance gear that they didn't have. I told the
agent that Joe Orsak had switched offNAMPS on cell 19 the pre-
vious night and that I thought that Cellular One's site didn't have
the technology. Before I hung up I put them in touch with the
Sprint guys, who could give them other, more detailed informa-
tion about the cell sites they'd be dealing with.
   Sometime after 5 P.M., I reached Levord, who had recently
arrived at the Bureau's Raleigh office. He was trying to find a
    282                        PURSUIT
   place to stay for himself and the Quantico team, whom he had
   just heard were coming, and he sounded irritated to be playing
   travel agent. ""Why are you having to mess with this, Levord?" I
   asked, sympathizing.
      He didn't answer.
      Without revealing to Levord that a foulup on our side of the
   investigation was now making me more anxious than ever, I
   observed that it looked as if we were going to need a much big-
   ger team of agents if a stakeout and arrest were to take place.
      "We're not going to get any more agents tonight," he said matter
   of factly.
      "Look, we've got to be ready to move tonight," I argued, but he
   seemed unwilling to do anything unless he was in control of the
   situation. Nevertheless, we agreed to meet at 8 P.M. at the Sprint
   switching office, where we could hook up with Murph and
   Orsak, and then have dinner nearby while we waited for the two
   Quantico agents.
      At 7:30, just as Julia and I were preparing to leave the hotel,
   Seiden called, sounding worried. Mitnick had returned to
   Internex less than an hour before, and it was evident that he knew
   something was up. "Looks like he's added an account called
(" Nancy, deleted Bob, and changed a lot of passwords-including
I  mine and root's," Seiden said. "This looks vindictive. He's getting
L  destructive now." And, in a show of spite, Mitnick had made
-" Markoff's account accessible to anyone on the Internet.
      "When I called to check in with Andrew, he said that he, too,
   had watched Mitnick's session on Internex, and that Mitnick was
   clearly acting paranoid. Mter leaving Internex, Mitnick had next
   gone to check his back door on Netcomsv that John Hoffman
   had closed on Friday. It was only one of several of Mitnick's ways
   into Netcom, but finding this particular entry barred he now
   seemed truly suspicious.
      His next action, according to Andrew, was to head directly to
   another Internet site we hadn't seen Mitnick use before, operated
    by the Community News Service in Colorado Springs, where he
   had a spare copy of test! salted away. This was the program which
   allowed him to use Netcom as a base of operations without leaving
   a record. It appeared that Mitnick brought back this fresh copy of
    test! to compare to the one he already had squirreled away on
                          THE STAKEOUT                           283
Nercorn, presumably to see if we had doctored the Netcom ver-
sion so that it might no longer hide his tracks. Comparing the
two copies, he found the Netcom version intact. He was using an
account named Wendy on Netcom with a password "fuckjkt."
   "Who's jkt?" Andrew asked.
   "I have no idea," I said impatiently.
   Andrew then described a series of activities that were fairly rou-
tine, by Mitnick's standards, which indicated to us that once he
had verified that his copy of test! had not been tampered with, he
had begun to calm down, perhaps concluding that the one barred
back door was a fluke having nothing to do with his problems at
Internex. Or so we hoped-at this stage in the game it was becom-
ing hard to tell what was calculated and what was coincidental.
After a few minutes, Mitnick had headed back to Internex, and
Andrew stopped watching him. We could.tell Mitnick was trying
to see if he had been detected and if so, where.
   "He's still on, so that's good," I said to Andrew. "But he's suspi-
cious. That's exactly what we don't need. After all the prodding I've
been doing to get the Triggerfish team here, it would be really
embarrassing for him to go radio silent for a week."

Markoff was hesitant about accompanying Julia and me to the
Sprint office. He was certain that when the FBI discovered a New
York Times reporter on the scene they would go nonlinear.
   "Don't worry about it," I said. "Just tell them you're on our
team. "
  "There's no way I'm going to lie to them," he replied. "That
kind of stuff always blows up in your face." But he was not about
to miss the story, so he decided to come along, bringing his own
car in case he had to bailout at some point.
  As Julia and I drove in my rented Geo Metro, I received a page
from a local number I didn't recognize. When I called it, I dis-
covered it was the home phone ofJohn Bowler, the assistant U.S.
attorney whom Kent Walker had promised to call.
   "Kent said you needed help," Bowler said. "What can I do for
   "I'm on a cell phone," I warned him.
   "Oh ,0k ay. "
284                         PURSUIT
   "There's someone coming down from Washington," I said.
   "Tell him to call me," Bowler replied, and we quickly ended
the call.
  Joe Orsak and Murph were waiting for us at the Sprint office,
joined by a third technician who was even larger than they
were-Fred Backhaus, a burly man with an unkempt beard, a
ponytail, and a motorcycle jacket. Despite looking like a Hell's
Angel, he turned out to be as friendly and hospitable as the oth-
ers, and all three were eager to get on with the chase. We talked
cell phones for a while, until Levord Burns arrived.
   Special agent Burns was an athletically trim black man with his
hair cut military short, whom I guessed to be in his late thirties.
His well-tailored gray suit, starched white shirt, Rolex-style
watch, and black wingtips would have put him in good standing
on Wall Street. But his big Ford Crown Victoria, with its omi-
nous whip antenna, looked like nothing but a cop car-a cop car
with Virginia plates. Look out, Kevin, I thought, the Feds are in
   The three engineers and I introduced ourselves, and I gave
Levord the message to call John Bowler at the local U.S. attor-
ney's office. He nodded noncommittally, not thrilled to have yet
another person to be accountable to.
   Burns told us that his bosses in Washington had ordered him
to bring along a set of Clipper phones-the government's new
standard-issue devices for encrypted digital conversations over
regular phone lines. They were in the trunk of his car. "In fact,
they're useless unless you're talking to someone who has one on
the other end," he said, rolling his eyes.
   "The trunk's a fine place to leave them," I agreed. Levord
might turn out to be okay, after all.
   Before we left the Sprint parking lot I introduced him to Julia
and Markoff, using only their first names. Levord asked no ques-
tions, and I offered no explanations.
   We drove in three cars to Ragazzi's, an Italian restaurant about
two kilometers from the Sprint switch. As we all sat down at a
single long table, I noticed that Markoff had picked the seat far-
thest away from special agent Burns.
   The restaurant was all done up with Chianti bottles and braided
garlic, but the breadbaskets were plastic. And while the bread-
                         THE STAKEOUT                         285
sticks were freshly baked, the salad turned out to be strictly ice-
berg. During dinner Levord talked a little about how the FBI
now routinely tracked cellular phone calls during investigations.
He acknowledged that they usually monitored people who didn't
know anything about the technology they were using-not peo-
ple who were cell-phone savvy like Kevin Mitnick. As he dis-
cussed his work, it was clear that Levord Burns was a guy
stretched pretty thin by a heavy caseload. "This kind of travel
puts a lot of stress on your family life," he added. "My wife is
pregnant, and I'm not around a lot."
   The Sprint guys followed suit and told us about their jobs, and
gave more details about the fraudulent cell phone ring they had
taken down. The raid had taken place at a farmhouse whose liv-
ing room had no furniture, just lots of cell phones on the floor.
The conversation turned to telephone fraud generally, and
Markoff recounted some of Kevin's history of manipulating the
phone system and how he had last been seen running from a copy
shop in Los Angeles.
   At one point Levord went off to a pay phone to return some
pages. -while he was gone we moved to the topic of Mitnick's
social engineering, and I recalled how he had tried to social-
engineer me at Los Alamos.
   "We've had a problem like that just in the last couple ofweeks,"
Murph said, surprised. "Somebody called one of our marketing
guys pretending to be a Sprint engineering employee, and he
managed to talk the guy out of several MIN-ESN pairs."
   "You don't happen to remember what name the caller used?" I
   Murph turned to Joe. "Do you remember?" Neither did.
   "Was it Brian Reid?" I offered.
   "Yeah, that was it," Joe said.
   "Kevin!" Markoff and I said in unison.
   -what an amazing creature of habit to stick to the very name he
had used on me severalyears earlier.The real Brian Reid was now
an executive running DEC's Internet networking business.
   The Sprint technicians were clearlychagrined to learn that Kevin
had weaseledinformation from their company. It wasn't their fault,
but it was a point of honor with them that they ran a secure shop,
and they were newly irritated by their colleague's lapse.
        286                        PURSUIT
          The more our conversation focused on Mitnick, however, the
        more nervous I became. If we'd had good operational security,
        we wouldn't have been having such a discussion in a public
        restaurant. I looked behind me and noticed a middle-American-
        looking couple sitting in a nearby booth, obviously interested in
        us. This made me even edgier. I began asking the Sprint guys
        technical questions to steer the talk in another direction.

        We had been back from dinner less than twenty minutes when
        the two-man team from Quantico finally showed up at the
        switching office, driving an old station wagon loaded with gear.
           They looked more like Simon and Garfunkel than G-men. One
        was tall with a rather sallow complexion, and the other was short
        with a fleshy nose and ears. Both appeared to be in their forties,
        and had a slightly rumpled look about them-tweedy academic
        with a touch of the racetrack-and the short one was wearing the
        type of cap favored by men who drive British sports cars.
           After introductions were made we decided that the best plan
        would be to use Fred Backhaus'swhite family van to transport the
        team and their direction-finding gear. As they began to unload
        the station wagon, Levord went off to the restroom to change
        from his business suit, emerging in work clothes and a baseball
        cap, an outfit that made him look like a house painter, though a
        painter a bit thick around the middle, thanks to the soft body
        armor he wore under his clothes.
           I volunteered to go along, because none of them had any idea
        of what the terrain was really like. Levord studied me, and in his

        slow, deadpan style said: "Tsutornu, there's no way you can come
        with us. Your picture's been everywhere. If he sees you and rec-
        ognizes you, he'll disappear."
           I persisted, though I knew I wasn't endearing myself to any of

        them. "Look, I should be along," I argued. "We made promises
        to all these computer sites that we'd be looking out for them.
        Nobody knows how Kevin might react. If he does something bad
        before you grab him, I need to see what's on his computer, so I
        can tell my people how to counter it. Until then, I can stay out
"       o f si ht. "
           Levord was unmoved. "Nothing's going to happen tonight."
                          THE STAKEOUT                           287
I got the clear message that he wasn't going to want me around
even when he did think something was going to happen. I sus-
pected that he felt intimidated by all this technology that he did-
n't understand, as well as not wanting to get blamed if I wound
up in a chase or a shootout.
   The Quantico team did talk to me a little about the technology
they had toted along in the station wagon, especially something
called a cell site simulator, which was packed in a large travel case.
The simulator was a technician's device normally used for testing
cell phones, but it could also be used to page Mitnick's cell phone
without ringing it, as long as he had the phone turned on but not
actually in use. The phone would then act as a transmitter that they
could home in on with the Triggerfish directional antenna.
   Clever as the technique sounded, I pointed out that it might be
risky to use on Mitnick. "You're dealing with someone who has
source code for all sorts of cell phones," I said. "He might be able
to detect it."
   They conceded that it might not be worth the risk, while
adding an unstated Go away kid, you're bothering us. I don't think
they liked the idea of dealing with a civilian, particularly one who
was in a position to learn all about their techniques.
   Backhaus had by now backed his van up to the front door of
the Sprint building, and the agents began moving back and forth
between their station wagon and the van, installing their gear.
The Triggerfish direction finder, a rectangular box of electronics
about a half a meter high, controlled by a Macintosh Powerbook
portable computer, was placed in the center of the van's back seat.
From one of the agents, who was sitting in the van calibrating the
unit, I was able to extract that the Triggerfish was a five-
channel receiver, able to monitor both sides of a conversation
simultaneously. Next they strung a black coaxial cable out the
van's window and ran it up to the radio direction-finding anten-
na they had placed on the roof. The roof unit had a black base,
about 30 centimeters square and several centimeters thick, which
held four long silver antenna prongs, each nearly 30 centimeters
high, reaching skyward.
   This apparatus seemed none too subtle, and I pointed out
again that they weren't dealing with some technically illiterate
 cocaine dealer. "This guy's paranoid, and he's been known to use
288                         PURSUIT
scanners to monitor the police before," I said. "He's wiretapped
the FBI in the past."
   They didn't want to talk to me at all now, but I wasn't going to
give up. "No, this is ridiculous," I said. "You guys are going to
park out there, and he's not stupid. I'm sure he knows what a
direction-finding antenna looks like."
   They didn't buy it. "It's not that visible," the short agent
   I looked at it ruefully. "Can't you put it inside?"
   "No, that would degrade the performance," the taller one said.
   "Why don't we put a box on top of it?" Murph suggested.
   "No, that would be too obvious," he said.
   I looked again at the top of the van, which had two parallel
rails running across it from side to side, as a carrying rack. What
we needed was a box that looked as if it was meant to be carried
   "Wait a minute," I told them. "Murph, you have fluorescent
lights. Do you have any of the boxes they come in?"
   We were in luck, they were in a storage locker in an equipment
room off the switching center's main room. We came back out
with a two and a half meter long box that could be lashed on top
of the van. I cut a hole in it so that it could be placed over the
antenna, completely hiding it in case Mitnick was in an upper-
floor apartment and might see the van from above.
   After we were done lashing and taping the box, the vehicle
looked like a respectable electrician's van. I could tell that the
agents had agreed to the camouflage mainly to humor me, but
they had to concede that the disguise worked pretty well.
   It was nearly midnight when the three FBI agents were ready
to roll.
   "So what happens if we actually see him outside his apartment?"
one of the Quantico team asked. It seemed probable that Mitnick
would frequent the shops in the strip-mall shopping center across
the road from his apartment complex. "Do we grab him?"
   "He's a probation violator, so we can take him in," Levord said,
"but would any of you recognize him on sight?" The photos that
all of us had seen were old, and the FBI documents indicated that
his weight had fluctuated.
   We decided that it seemed unlikely they would get further
                         THE STAKEOUT                         289
tonight than simply identifying which apartment was his, so the
Quantico team left with Orsak and Backhaus, while Joe and
Levord followed in my green rented Geo, which they decided was
the least suspicious vehicle in our fleet. Levord said they would
do a quick surveillance and be right back.
   While we waited Murph gave us a tour of the switch, a win-
dowless building filled with equipment that looked a lot like
mainframe computers, racks of dumpster-sized batteries, and
standby generating equipment. Then we settled in to wait-for
minutes, and then, to our growing anxiety, for hours.
    A little coffee room off the main operations center gave us a
place to sit, and we passed the time eating crackers and drinking
soft drinks from a small refrigerator. A handmade sign listed
prices for different items, including Gatorade, ofwhich there was
none left, unfortunately. Payment was on the honor system, and
the large jar that had been provided for that purpose was slowly
filling with our dollar bills. To keep occupied, I studied the bul-
letin board with all of the relevant OSHA notices, and read a
newspaper clipping about the farmhouse cell phone bust that
Murph and Joe had been involved in.
   Markoff and Julia began to play with the HP 100 that was part
of my RadioMail terminal. At one point Markoff started a pro-
gram that was supposed to edit icons for the device's user inter-
face, which somehow caused the computer to abruptly crash, cor-
rupting all my wireless communications software.
   Markoff apologized profusely, but the computer's communica-
tions system was quite dead. I remembered that all the backup
files were safely in San Diego, where they would do me little good
now, which brought to mind a quote I'd once read, from some-
one I couldn't recall: "Fate is infatuated with the efficient."
   Mitnick was not using the Sprint system. And while we could
find out that he was active on Cellular One by calling in period-
ically to check with Gary Whitman, a Cellular One engineer who
was monitoring the cell site from home, we could not follow
Mitnick's calls as closely as would have been possible if they were
being switched through the building in which we now sat.
   At around 3 A.M. I paged Joe Orsak. He quickly called back,
but he couldn't tell me much, other than to say he was calling
290                        PURSUIT
from one of the pay phones at the strip mall across Duraleigh
Road from the apartment complex.
   I asked him to please call one of the FBI agents to the phone.
A couple of minutes later, one of the Quantico agents came on
the line, and without waiting to hear what I wanted, asked furi-
ously, "Who's this guy named John who's with you?"
   "He's a writer," I replied.
   "What does he write?"
   "He's a writer. He writes books."
   "Does he write anything else?"
   "Lots of stuff." I thought I understood his problem-he'd be in
deep trouble with his superiors if he had knowingly let a news-
paper reporter observe the team's activities. I was trying to give
him the option of plausible deniability, but he pressed on.
   "He wouldn't be John Markoff, the New York Times reporter,
would he?
   "Yes, that's him," I was forced to concede.
   "And did he write that book about computer hackers?"
   "Yes. Cyberpunk. The book on Kevin Mitnick. He's our
Mitnick expert."
   Now he was really enraged. "Why is he here? Why is he along?"
he asked. "You're endangering the operation! Reporters are not
allowed to take part in FBI activities! You lied to me!"
   "No," I answered. "I never lied to you. You never asked who
he was."
   My explanation didn't satisfy him, and he hung up.
   Markoff had heard my end of the conversation and decided it
was time to make a quick and graceful exit. He had made his
identity perfectly clear to Joe and Murph the night before, even
exchanging business cards with them, and one of them had
apparently mentioned it to the FBI men.
   "I don't want be caught in the middle of this and have to
explain my presence to an FBI agent," Markoff said before he
drove back to the Sheraton.
   Forty-five minutes later, at nearly 5 A.M., Levord Burns
returned with Joe Orsak.
   Levord, forgoing his customary slow-motion approach,
stormed into the room and tore into me as soon as he saw me.
"Look, you've been jerking me around, and now I find out you
                         THE STAKEOUT                          29 I
have this New YOrk Times reporter shadowing us. What's going on
  "Let's talk," I said. I saw that I needed to peel Levord off the
ceiling, and I didn't think it was going to do any good for him to
lay into me while Joe, Murph, and Julia looked on. I motioned
toward the storage room.
    We closed the door behind us, and Levord started pacing back
and forth. "What are you trying to do?" he said. "What's your
agenda here?"
   I told him I wasn't trying to grandstand by bringing along a
reporter, but merely relying on a friend I trusted, someone who
had been writing about Mitnick for many years and had some
expertise on his habits and motives.
   It was clear that I had run afoul of Levord's FBI obsession with
operational security, and he let me know that the Quantico
agents were terrified that the secrets of their surveillance proce-
dures were going to show up in a New YOrk Times article. The
agents were going to have to report the encounter to their supe-
riors, and they were very unhappy about it. Although Burns
looked exhausted, he lectured me for a full twenty minutes about
the Bureau's specific protocols for dealing with the press, and said
that many had been violated by us.
   "Is he going to tip Mitnick off so he can escape and make it a
better story?" Levord asked.
   "No way," I assured him. Obviously, as a reporter, Markoff was
here because it was going to make for a good story, but the best
story would be if Mitnick got caught. So Markoff's interests and
Levord's were completely in synch, I explained to him.
   "Why didn't you introduce him as a reporter? Why didn't you
come clean with me?"
   "You never asked me," I replied.
   "Where is he now?"
   "He went back to his hotel."
   Finally, as Levord's rage seemed to peter out, I asked him if
they'd nailed down Mitnick's location.
   "We're close," he said gruffly, "but we still haven't pinpointed
the exact apartment." The Quantico agents were still trying.
   Although I was fairly certain that he still would have liked to
throw me out, Levord probably knew that right now was when
292                         PURSUIT
he needed me most. He was going to have to put together an affi-
davit for an arrest warrant. And that would require much of the
data we had collected so far, which he had no way of interpreting
without my walking him through it.
   With that in mind, I offered to help.
   "Okay," he said, "but no more surprises-right?"
   I nodded, then suggested the types of data I could correlate for
him: Netcom log-in records, the cellular company records, and
Mitnick's actual sessions that we had monitored. Cross-referenc-
ing these would demonstrate irrefutably that Mitnick was our
man. I called Andrew and asked him to fax me some more of the
relevant material.
   As Levord and I headed for the coffee room to get started, Julia
came up, saw that we had apparently reached a cease-fire, and
then slipped into the storage room behind us to take a nap.
   "Sorry," she said. "I've got to get some sleep."
    Andrew's backup material started rolling off the fax machine a
short while later. For the next hour and a half, I sifted through
Andrew's data and mine, while Levord spent most of the time on
the phone, briefing various Bureau officials and making arrange-
ments for backup and reinforcements.
   I wound up with a listing of thirty separate sessions, which
took place between the afternoon of February 9 and the early
morning of February 13, for which we could match the Netcom
log-in times with Cellular One or Sprint call-detail records. From
these I selected a handful for Levord, and began explaining how
the phone records and Netcom sessions corresponded, and what
Kevin's actual keystrokes for each session told us about his activ-
ities. For anyone without a good grounding in phone networks,
and Internet and Unix commands, it was a daunting amount of
information to digest; given Levord's and my relationship at this
point, the process was particularly painful. But soon Julia awoke
and took over, proving to be a much more patient instructor.
   When Julia and Levord wrapped up, she and I stepped outside
for some fresh air, having been holed up all night in a fluorescent-
lit, windowless building. I was surprised: daylight, a leaden, over-
cast sky, but morning all the same. It was nearly 8 A.M.
   Back inside, I called Andrew and Robert Hood, who were still
waiting at Netcom in case something happened, and sounding
                          THE STAKEOUT                           29 3
grumpy about it. I told them that Levord was now saying that the
FBI would probably go out and pick up Kevin before noon
Eastern time, as soon as his affidavit and the arrest and search
warrants were complete. The "get set" and "go" signals might be
coming to them in the next few hours, I assured them.
  A little later, the two Quantico agents returned, with the hag-
gard appearance of two middle-aged men who had been up all
night. They gave me a baleful look, but neither they nor I had the
energy for a confrontation about Markoff. Agents from the
Raleigh office had just relieved them, and now they intended to
check into their hotel and catch some sleep while Levord chased
the paperwork. He was heading downtown to the Bureau office,
and because the monitoring had still not produced a precise
apartment number, he wanted to have someone go by the Players
Club complex for some old-fashioned, discreet legwork. I got
Levord to promise he would contact me before going out to make
the arrest. I didn't believe he'd keep to it, but I kept stressing the
vulnerability of the Well and Netcom and other sites and the
need for me to alert the people there.
  The cellular telephone switching center was starting to come
back to life, as the small crew of workers arrived for their day
shift. Julia and I headed back to the Sheraton in the Geo, and she
pointed out that it was February 14-Valentine's Day.

I woke with a start. The curtains were drawn, and I had to roll
over and look at the clock to see the time: nearly 2 P.M. I grabbed
my pager from the nightstand and pressed the backlight: no new
pages. Shit. They must have gone to arrest Kevin without notify-
ing me. The bust had happened, and Andrew and Robert were
probably sound asleep.
  I fumbled through my belt pouch, found the note with the
scribbled number of the local FBI office, and dialed.
  ''I'm trying to reach Levord Burns," I said as soon as someone
picked up.
  A sleepy voice on the other end of the line said, "Uh-huh." It
could only be Levord.
  "What's going on with the warrant for Mitnick? Have you
raided yet?"
294                         PURSUIT
  Instead of getting an answer, I listened as it sounded as if my
call was being transferred. Then the line went dead.
  Whom had I just been talking to? Had I been conned?
  Now it was my turn to be paranoid. What if Kevin Mitnick
had been able to tamper with the FBI's phones and had forward-
ed their calls to himself? If it was Mitnick, I had just given away
everything. I immediately called back the same number, and a
different male voice answered: "FBI."
  I asked for Levord Burns.
  "Who?" the voice said.
  "Levord B " I sal. "He'down fjrom ",VT. hi
              urns,     id        s            was mgton. "
  "I don't think he's here, but there must be thirty-five people
rushing around. It's pretty busy." I thanked him and hung up.
    Maybe the raid was going down now? By now Julia had awak-
ened, and I told her, "I think we should go back over and see if
anything is happening at the apartment complex."
  The Player's Club apartments were on the other side of the air-
port from the Sheraton, and Julia drove us there through what
seemed an endless maze of road construction and detours. When
we finally reached the complex we drove past it once, but I saw
nothing that looked remotely like a surveillance operation, a
stakeout, or the flurry of activity one might expect if a Federal
arrest had taken place during the last few hours.
   Not wanting to hang around the area any longer and risk being
seen, we drove a short way down Glenwood Avenue toward
Raleigh and found a pay phone at a gas station.
   I calledKent Walker, who said he had not heard from Raleigh that
day. It appeared that nothing at all had happened yet, which sur-
prised him and totally baffiedme. At Kent'ssuggestion, I calledJohn
Bowler, whose message I'd passed on to Levord the night before.
   "I've had no news," Bowler said. "I just had this dropped in my
lap. But I haven't seen any paperwork and I haven't heard from
special agent Burns." Despite being in the dark, Bowler sounded
willing to be helpful.
   "I think we need to talk as soon as possible," I said.
   Bowler gave us directions downtown to the Federal Building.
The afternoon traffic slowdown had already begun, so it took us
a while to wend our way down to the courthouse. We parked on
the street in front of the boxy, glass-front modern building and
                         THE STAKEOUT                          295
walked in, passing the guard stand in the lobby.
   Shortly past four o'clock Julia and I finally reached the u.S.
attorney's suite up on the top floor, where we signed in and
received visitor badges. We had to wait for a while for Bowler to
finish a meeting, before he came out to the reception area, intro-
duced himself and invited us into his office.
   The prosecutor, a balding man in his early forties, had a toothy
smile and a rosy-cheeked, almost mischievous demeanor. He car-
ried himself like an athlete and it was obvious he was something
of a bicycling fanatic, for biking magazines were scattered around
the office and there was a framed cartoon about the funny clothes
cyclists wear. There were also several family pictures of his wife
and his two preteen sons.
   We sat down in two chairs in front of his desk and began to
explain our reason for turning up at his office late on a dreary
Tuesday afternoon.                                           .
   "How much of this do you already know about?" I asked.
   "Very little," Bowler said, but he seemed intrigued that two
California computer hackers had wandered into his office with a
tale to tell.
   I told him that we were pursuing Kevin Mitnick, who was
wanted by the FBI and the u.S. Marshal Service, and I gave him
as precise a rundown as possible of the events of the past weeks,
up through tracing Mitnick to the Player's Club apartment com-
plex on Sunday night.
   "The FBI has been in town since last night," I said, "and since
we all now know where Mitnick is, I don't understand why things
aren't moving more quickly. He's managed to elude the Bureau
for more than two years, and it looks as if they're giving him every
opportunrty to get away agam. "
           .                   .
   "Is he armed, or is he dangerous in any way?" Bowler asked.
   I said I doubted that he was armed, but that he was dangerous
in unpredictable ways. Whether or not he would actually wield
that power, at the moment he was in a position to damage com-
puter systems used by tens of thousands of people and containing
property worth hundreds of millions of dollars. Several Internet
companies were operating at considerable risk in an effort to help
us catch this criminal, and were not likely to keep exposing them-
selves much longer.
29 6                        PURSUIT
   "Mitnick isn't your ordinary criminal," I emphasized. "This is
a game to him, and he understands telephone and computer
technology a lot better than the law enforcement agents who are
pursuing him."
   I told Bowler the special agent on the case, Levord Burns, had
expected to have an affidavit for search and arrest warrants by
noon, but I hadn't heard from him since early morning. I was
concerned that this not carryover to yet another day, because we
now had reason to believe that Mitnick might be on to us.
   "Sounds like I need to talk to Levord Burns," Bowler said. He
called the Raleigh FBI office, which referred him to Levord's hotel.
   When he reached Levord in his room, Bowler said, politely but
firmly, "I understand you're working on the affidavit." He paused
and listened for a moment. "Can you get over here as soon as pos-
   Bowler looked at his watch-it was nearly the close of the busi-
ness day in the Federal Building-and said, ''I'd better line up a
judge for these warrants." He placed a call to the office of
Magistrate Judge Wallace Dixon, whom he subsequently tracked
down at the building's fitness center. Bowler and the judge
arranged for us to bring the paperwork by his house later.
   Bowler's next calls were to a friend, who agreed to sub for him
that evening as coach of his son's soccer team, and to his wife, say-
ing he'd be missing the game and would probably be a little late
getting home.
   Soon he was busy pulling together documents and delegat-
ing tasks to two assistants. The younger of the two women,
who looked to be in her thirties, was plump, with curly blond
hair, well-polished nails and a Betty Boop scarf. The older
woman, who also seemed to be more authoritative, was plain-
er, with the husky voice of an inveterate smoker. As they were
working, I received a page from Pei at the Well.
   "Something new has happened," she reported when I reached
her. Kevin had destroyed some accounting data, she said, and
while they had been able to recover it, the Well officialswere con-
cerned that he might have suddenly turned vindictive and would
now aim to do permanent damage. "Tsutomu," she said, "our
management is worried about leaving ourselves vulnerable like
                        THE STAKEOUT                          297
   "Management" sounded to me like "Claudia," so I gave Pei a
status report, concluding, "We know where he is and we're going
through the paperwork for the arrest right now." I told her I
would call Bruce Katz as soon as I got a chance.
     Hearing this exchange seemed to further galvanize Bowler,
who once again called Levord, telling him more firmly and a bit
less politely this time, "We really need that affidavit!"
   I was grateful to Kent Walker for having hooked me up with
Bowler, so I called to give him an update. He, too, was pleased to
hear that someone in Raleigh finally recognized the urgency of
the case, and was dumbfounded that Levord still hadn't produced
the necessary documents. "What's going on with this guy?" Kent
said. "We don't need to write a book on the subject. The affidavit
doesn't need to be that involved."
   Because Julia and I had not eaten since dinner the previous
night, and the Federal building was about to close, she went
downstairs and found a Subway sandwich shop and brought back
some sandwiches. We sat on the floor of Bowler's office and ate,
sharing our food with Bowler and his assistants, who were now
working through dinner.
   I phoned Katz, who recounted what Pei had told me about the
deleted accounting file. "Tsutomu, I want your advice," Katz
said. "How vulnerable are we?"
   Katz raised a series of questions. Had Mitnick actually figured
out that the Well's staff were watching him, and had decided to
take them with him if he himself was about to be taken down?
What were they risking by not shutting their systems down or
locking him out immediately? "What's going on here, Tsutomu?
Is he trying to get revenge?" Katz asked.
   "We haven't done anything to turn Mitnick against the Well,"
I answered honestly. "We're this close to getting him," I said.
"Give us a little more time."
   While I didn't think Mitnick had any reason to believe the Well
was on to him, I couldn't say the same for Netcom. I phoned
Andrew, who reported further signs of paranoia from Mitnick.
He was continuing to move his data stashes and change pass-
words, and as a gesture of contempt for all who cared to review
the log files, he had attempted to log into Netcomsv with the
password .fukhood, no doubt for Robert's special attention. And
29 8                        PURSUIT
unfortunately, there was also an indication that Mitnick was sud-
denly approaching the Well with new wariness: the dono
account, which he had been using for weeks with the same pass-
word, fucknmc, now suddenly had a new one. There may have
been some hidden meaning in the choice of dono's new pass-
word-no,panix-but what mattered far more to us was that
Mitnick had apparently felt a need to take a counter-security
measure at the Well, even though it turned out to be an ineffec-
tual measure, given the level of our surveillance. Had something,
or perhaps someone, tipped him off? Had he discovered Pei's use
of his log-in?
   Levord finally arrived, glaring at me briefly as he walked in,
and seeming to move even more glacially than his usual deliber-
ate pace. He placed the affidavit on Bowler's desk and said that
he might have been there sooner but he thought it might be a
good idea to organize his team if there was to be a proper stake-
out and arrest.
   "I've also taken some extra time to make sure that the affidavit
is correct. It might be your intent to catch him," he said, eyeing
me, "but it's necessary to do everything by the book in order to
keep him."
   The continuing obstacle, Levord told Bowler, was trying to
determine the correct address from among the several possibili-
ties to which the Quantico team had narrowed it down the pre-
vious night. The layout of the buildings was making it difficult to
get a precise fix on the cellular radio signals.
   That morning, the local FBI agent, L.B. Thomas, had gone to
the Player's Club and talked to the manager in hopes of win-
nowing their list by determining whether any men in their early
thirties had recently moved into one of the suspected apart-
ments. Two new tenants had arrived during the last two weeks,
it turned out, but one of them was the manager's girlfriend, and
the other lived on the wrong side of the complex. So Levord was
still left with a list of three possible addresses, any of which
might be the right one. There was also a chance that none of
them was the apartment from which the radio waves were actu-
ally emanating.
   The difficulty, Bowler explained to Julia and me, was not in
obtaining an arrest warrant, which could be issued for the entire
                         THE STAKEOUT                           29 9
Player's Club complex as long as Mitnick was somewhere on the
premises. The sticking point would be the search warrant. To
look for and seize evidence, it was necessary to have a judge's per-
mission to search a particular residence, which in this case would
have to be specified by building number and apartment unit.
   Levord went out to the reception area to put out more calls to
see if his team had come up with any new leads. Meanwhile,
Bowler and his assistants, now that they had the affadavit to work
from, set about preparing warrants for each of the addresses on
Levord's list. For good measure, Bowler had them draw up a
fourth set, with the address blank. He hoped he could persuade
the judge to sign the three that were completely filled out, and
leave the blank one to be authorized later, if necessary, with a
phone call to the judge, if the agents on the scene should deter-
mine that Mitnick lived in some other apartment.
   I helped Bowler compile a list of items to include in the search
warrant, such as computers, hardware and software documenta-
tion, floppy disks, modems, cell phones, and cell phone parts. It
was rather eerie trying to picture Mitnick's lair and what it might
contain. Word of his latest machinations at Netcom and the Well
had once again underscored his potential for maliciousness. And
when one of Bowler's assistants went to print the warrants, only
to discover that her computer was suddenly unable to talk to the
printer over the office's local area network, Julia suggested that it
might somehow have been Mitnick's handiwork. But we quickly
discovered that the problem was a bug, not a saboteur.
   Finally, shortly after 7 P.M., we had the warrants assembled.
Since Levord had gotten no further on narrowing the address list,
Bowler bundled up all four packets and we headed off for Judge
Dixon's home.
   As we walked through the anteroom outside Bowler's office
Julia noticed a jar full of small candy Valentine hearts. She poked
through it and found one that had YES DEAR written on it and
handed it to me. She meant it as a joke because I sometimes
teased her by saying that, but I was so distracted that I only
looked at it absentmindedly before popping it into my mouth.
The four of us headed out of the suite for the elevators, cheered
on by the two assistants.
   "Go get 'em!" called the one with the husky smoker's voice.
300                          PURSUIT

We decided to ride to the judge's house in Bowler's customized
van, so we could proceed directly to the Player's Club afterward.
I was still hoping to get to the apartment complex before 8 P.M.,
to catch Mitnick on the air before he signed off for his dinner
break, which usually lasted until 11 P.M. or so. The van, outfitted
with louvered windows and curtains, would make it easier for me
to stay out of sight during the stakeout. It would also be more
comfortable than the Geo, since it was a mobile family room,
complete with wood paneling, fuzzy upholstery, and food wrap-
pers and plastic action-figure toys on the floor.
  As Levord followed in his Crown Victoria, we drove to an
affluent north Raleigh suburb which turned out to be quite near
Mitnick's place. Julia and I waited in the van while Bowler and
Levord went into the judge's house, a modest-sized brick home
with a small, roofed porch. It was dark enough outside that we
could easily watch through the judge's uncurtained living room
picture window and see people moving about inside.
   While we waited I decided to send Andrew the ready code we'd
agreed upon, which turned out to take some doing. I wanted to
bracket the number 080663 with dashes, to make clear at a
glance that this was not a regular phone number. On many
numeric pagers the dash is created by punching the * key, but
when I entered the combination *080663* followed by the # key
to send it, I got a fast busy signal, indicating some sort of error.
After I tried it again, with the same result, I entered the code
number without the dashes, and after it was successfully trans-
mitted, hoped Andrew would interpret it correctly.
   Procedures were not going smoothly inside the judge's house.
Judge Dixon, as we learned later, was requesting various changes
in the warrants, including a proviso that an arrest could be made
after 10 P.M., a necessary legal stipulation, since arrests are gener-
ally supposed to be made during normal waking hours. It was
clear that additional documents were going to be needed, so
Bowler made arrangements for someone from his legal staff to
prepare them and meet up with him at the stakeout. Just before
8:30, Bowler and Levord emerged.
   We drove several kilometers to the shopping center parking lot
across Duraleigh Road from the apartment complex, and Levord
peeled off down a side street to the far side of the Player's Club,
                         THE STAKEOUT                         30 I
where the action would be. We had no idea at this point how
many or what sort of reinforcements Levord had assembled, but
I assumed that since this part of the show was totally in the FBI's
hands, they would know what they were doing.
   Bowler circled the small shopping center parking lot while we
considered how best to position the van. He finally parked so that
we could view the Player's Club complex, though not Mitnick's
side of it, through the windshield. I sat far in the back to be
inconspicuous. After setting up my monitor, I switched on a tiny
flashlight I had borrowed from Julia to watch the display as I
began searching both the Cellular One and Sprint frequencies.
There was no sign of Mitnick.
   "Looks like he's having dinner," I said. Maybe he'd gone out;
maybe he was even here in the strip mall. We peered around out-
side the van, but there was no foot traffic along Duraleigh Road.
Bowler and Julia decided to take a walking tour of the area to see
if they could see anyone who might be Mitnick, though neither
of them knew exactly what he looked like.
   For the chilly February night air Bowler was wearing a fedora
and a trench coat, and as he got out of the van he turned to me
and said slyly, "Do you think I look too much like an undercov-
er agent?" He was obviously enjoying this unexpected adventure.
   With the side curtains now closed so that I wouldn't be seen, I
couldn't tell what Bowler and Julia were up to, but they were back
within fifteen minutes with a reconnaisance report. As I listened,
Bowler let me wear his fedora so that I wouldn't feel left out of
the caper. He and Julia had passed a Chinese take-out restaurant,
a pizza delivery shop, and a bar, glancing through their windows,
then entered the convenience store to buy some Gatorade, pop-
corn, and AAA batteries for Julia's flashlight.
   "I didn't see anyone who looked like Mitnick," Julia said drol-
ly, "but there were sure some suspicious-looking characters across
the parking lot." On the way to the shops, they had walked past
two men somberly sitting in the front seat of what looked dis-
tinctly like a government car, facing the Player's Club.
   Julia had decided to go into the Chinese place to get us some
more food, and while waiting for the order to be filled she noted
that one of the counter people looked out at the same car and
said to a coworker, "Something must be going on; those guys
302                         PURSUIT
haven't moved for the longest time." One of the pizza parlor
employees had also ducked his head out the door several times to
see what might be taking place in the parking lot.
   For the next two hours I encountered nothing but silence on
the scanner, and I couldn't banish the thought that Kevin might
have fled, a scenario that seemed increasingly plausible, given the
lack of subtlety of what little we could see of the stakeout.
   "We saw him talking about seeing Sneakers the other night,
maybe he's gone to a movie," Julia said at one point, trying to
lighten my mood.
   A cluster of outdoor pay phones stood near the gas station in
the corner of the parking lot, which we could see by peering
through the side curtains. A remarkably large number of calls
were being placed from them for a weeknight in midwinter.
Several times we were paged by Markoff back at the Sheraton,
wondering if there was any news, and Julia would get out of the
car and walk to the pay phone to return them. Then, close to 11
P.M., there was a page from Andrew, at Netcom.
   "You're not going to like this," Julia warned, when she returned
from talking to Andrew. He had realized that he'd screwed up
again. Three hours earlier, when it had taken me several attempts
to transmit the "get ready" message, Andrew had interpreted the
flurry of signals to mean that Kevin had already been arrested.
For evidence, he had started to make backup copies of the files
that Mitnick had stashed around the Internet, and then begun
deleting the intruder's own versions.
   There was also one piece of good news-Andrew had analyzed
Mitnick's deletion of the Well's accounting file earlier in the day
and had determined that it was the result of a simple typo, not an
act of sabotage. But the bad news was devastating: our surveil-
lance had now been irredeemably compromised.
   And it had happened several hours ago. Again, Andrew had not
 called sooner, fearing my anger. This was unbelievable. Here I'd
been riding the FBI as hard as I could, and now if everything fell
 apart and Mitnick escaped they were going to be able to come to
me and say, "Your guys blew it."
   But there was no time to fret about the error now: my monitor
 indicated that Kevin Mitnick had just signed on for the night
 shift. And if he hadn't noticed before dinner that his stashes had
 been destroyed-and his presence now indicated he might not
I                            THE STAKEOUT                         303
I   yet know-he was about to find out.
      I wasn't the only one who'd heard Kevin come back to life.
    Suddenly Levord's car and several other vehicles sped through the
    parking lot and disappeared behind a bowling alley at the end of
    the shopping center. It was a quick, final coordinating meeting of
    the federal and local law-enforcement agencies, and Bowler drove
    the van around to join the half-dozen plainclothesmen who had
    assembled. He handed Levord the amended warrants, and I
    warned the group that Mitnick might have been inadvertently
    tipped off, so haste was more crucial than ever. Someone men-
    tioned the Quantico agents now had a "beacon" to home in on
    and could now use a hand-held signal-strength monitor for close-
    up work, so it shouldn't take long to find him. The meeting last-
    ed less than a minute, and the others were off to take up their
    assigned positions around and on the far side of the Player's Club.
       Bowler eased the van back to our parking spot, and I resumed
    monitoring, picking up Mitnick as he came and went for the next
    hour or so. Usually, upon concluding, or dropping, a data call, he
    would redial immediately. But at one point, I didn't hear him
    come back on, so I began checking the adjacent arcs of the
    Cellular One cell site, to see if his signal had been bounced to
    another sector. And then I noticed something strange.
       Although Kevin was south of the cell site, I was now picking
    up a data carrier from the north. It was the first time I'd seen
    another data call being placed within this cell from anywhere but
    Mitnick's vicinity since I'd come to Raleigh. Because of the spot-
    ty reliability of cellular connections, and the fairly high cost of
    the service if a person is not stealing it, it is not common to use
    cellular radio for transmitting data.
       I reported my discovery to Bowler and Julia, and we began
    conversing in hushed voices. Had Mitnick moved? Did he have a
    partner? Or had he gone mobile after being alerted by Andrew's
    premature cleanup?
       After I picked up the MIN of this new data caller, I told
    Bowler, "Drive over to the phones." It was 12:40 A.M. He posi-
    tioned the van as close to the pay phones as he could get it, with
    the vehicle between me and the apartment complex, and I slid
    out and called the Cellular One technician.
       "Gary," I said, when Gary Whitman picked up his phone. ''Are
    you watching?"
304                         PURSUIT
   He was indeed monitoring the Cellular One site, so I read him
the new MIN and asked him to let me know each time our mys-
terious caller placed a new call and moved to a different frequen-
cy. He could do so by paging me with the new channel numbers.
   Once again Bowler returned to our parking spot, and almost
immediately I got the first of a series of pages, which allowed me
to quickly flip back and forth between Mitnick's sector and the
mystery man's, confirming that we indeed had two separate data
callers using the cell site. For forty-five minutes, I continued
watching the two.
   Then, at almost precisely 1:30, Kevin's carrier went dead. We
immediately saw the Quantico station wagon zip past, first down
Duraleigh Road, and then a short time later speeding past in the
other direction. The car now had the directional antenna that
had been on Fred Backhaus's van the previous night. The second
data carrier was still on the air, and it was obvious that the
Quantico agents had spotted it, too.
   Other vehicles were now moving in on the Player's Club as
well-including our neighbors from the parking lot.
"Something's happened," Bowler said. "Let's go have a look."
   He slowly pulled the van out of the parking lot and onto a side
street nearer the apartment complex, stopping behind some
bushes just to the east of it, where we could see directly into the
parking lot. We stepped out of the van. We could now see that
the area was well staked out. There were at least four government
cars, and at least a dozen plainclothesmen standing or walking
around. The Quantico station wagon returned.
   I wanted to go tell the agents what I knew about the new signals
from the north, but Bowler came over. "No, no, no," he said, stand-
ing close to me. "There's nothing you can do at this point. Besides,
we don't know if they have Mitnick yet. He might see you."
   "But I've got the MIN," I objected. "Cellular One is sending
me the channel numbers."
   "Why don't you send Julia over with a slip of paper?" he sug-
   She walked over. They were annoyed at first, but when they
realized she had valuable information, they took the MIN and
my pager from her, and the Quantico teams' wagon roared off
once more, now heading north. Julia returned, and we sat in the
                         THE STAKEOUT                           305
van, listening to the soft hiss of the mysterious data caller's
  Ten minutes later, Levord strode over.
  "We're inside," he said. "We've got Mitnick. But we're going to
need to call the judge to authorize the warrant to search a new

I handed Bowler my cell phone so we could wake Judge Dixon for
authority to search Kevin Mitnick's home. Then I sent Andrew the
"go" code on my pager. Finally, it really was time for him to start
alerting the Well and the other sites on our warning list. I hoped he
was still awake.
   AB Bowler talked with the judge, Levord described for Julia and
me how he and several other agents had knocked on the door of
his apartment and waited a full five minutes for it to open. When
it finally did, the man inside refused to admit that he was Kevin
   They went in anyway, and Mitnick rushed to lock up some
papers in a briefcase-a futile act, under the circumstances. He said
he was on the phone, talking to his lawyer, but when Levord took
the receiver, the line was dead. America's most-wanted computer
criminal began vomiting on the floor of his living room.
   "We haven't Mirandized him yet," Levord said. "We're actually
worried about his health. We found prescription drug bottles.
He's on some sort of medication."
   Now it was safe to leave the van. The Quantico team had
returned, and I walked over to where they were standing next to
their station wagon. The clammy night air had finally given way
to a light rain. The agents had the air of weary veteran athletes
after a big victory-content, but far too tired to celebrate. The
radio echoes had continued to bedevil their directional antenna,
so they had finally tracked down Kevin by walking through the
complex with the signal-strength meter, watching the cellular sig-
nal grow stronger and stronger until it led them to Mitnick's
front door.
   I asked about the other data signal. They said they'd chased it
for a while but had been unable to pin it down. It, like many
other threads in the investigation, would remain a mystery.
306                          PURSUIT
   I never saw Mitnick that night. It would be at least another
hour before he was finally hauled off to a holding cell downtown
in the Wake County Jail. Long before then, Levord Burns
obtained a valid search warrant from the judge, via Bowler, and
had his agents start gathering up the evidence.
   Levord came over to the van again to report this progress. I
asked if I could have a look at the apartment, to see how my
opponent had spent his days and long nights, but he declined to
let me do so.
   "We've taken lots of pictures of the inside of his apartment, but
they're evidence for the trial, and no one else will see them until
after it's over." But he did point out Mitnick's car, an old, light
blue Plymouth Horizon.
   Levord's mood was noticeably improved, despite the steady
cold drizzle that was now soaking us all, and he came around to
the side of the van and shook my hand.
   "Congratulations," I said. "We managed to do this without
killing each other."
   He didn't reply, but for the first time since I'd met him, special
agent Burns smiled at me.
I was awakened the next morning by a call from Markoff.
    "Kevin's court appearance is at ten o'clock," he said.
    I looked at my watch and saw that it was already a few min-
utes past nine. "We'll meet you in the lobby as soon as we can
get dressed," I told him, gently shaking Julia awake, then pulling
the hotel room curtains back on a gray wet Raleigh morning.
    Markoff drove us downtown through a light rain, and because
I . was feeling a bit out of touch since the crashing of my
RadioMail terminal, I decided to use my cell phone to check my
voice mail in San Diego. I couldn't believe it. There was a new
message in that phony Asian accent, and it had been delivered
just before 7 A.M. West Coast time, a full eight hours after
Mitnick's arrest but well before any news of his capture had been
reported by the media.
    The message was long and rambling with none of the cock-
iness or bravado we'd heard before but, instead, in a delivery
so nervous and rapid that the accent occasionally fell away
altogether. After listening to it, I played back the message
twice more, first holding the phone to Julia's ear, then to
    "Hi, it is I again, Tsutornu, my son. I just want to tell you-
very important, very important. All these phone calls you
received with, ah, making reference to Kung Fu movies-noth-
ing to do with any computer thing whatsoever. Just a little, ah,
interesting call.
3a 8                          PURSUIT
   "I see now that this is getting too big, way too big. I want to
tell you, my son, that these have nothing to do with any com-
puter activities whatsoever. Just making fun of Kung Fu movies.
That's it. That's it.
   ''And making reference to, ah, you know, trying to make a refer-
ence to putting Kung Fu movies into the ... into a computer ref-
erence. That's it. Nothing to do with any Mitnick, hacking, any-
thing, nothing. I tell you it was just a interesting call that's ... it.
All coincidence. This is getting too big, and nothing wrong has
been done by anybody who left any messages on your voice mail.
Just to let you know. Okay? It's getting way too big."
   We were amazed. "So the tables have turned," I said. I won-
dered aloud where Mitnick's friend had gone to ground. I was
curious if he was hiding right here in Raleigh. Whom had
Mitnick called in the minutes before he opened the door for the
FBI the night before? Was this the owner of the second cellular
phone that had been making the data calls that the Quantico
team had been chasing?
   We were still talking about this new mystery, as we entered the
Federal building. It was just a brief prearraignment hearing, and
word hadn't yet gotten out that Kevin Mitnick had been arrested.
We walked into a small empty courtroom and sat down in the
last of the three short rows that had been reserved for spectators.
It was like U.S. courtrooms allover the country, an austere, win-
dowless space with a high ceiling.
   After a short time Mitnick was led in from a door at the front
of the room to the right of the judge's dais by a dour U.S. mar-
shal. Kevin didn't look ill, but he also didn't look anything like
the overweight, bespectacled "dark side hacker" who had once
terrorized Los Angeles. We saw a tall young man, neither thin nor
stocky, who had metal-rim glasses and shoulder-length flowing
brown hair. He was wearing a gray charcoal sweatsuit, and he was
handcuffed and his legs were chained.
   Halfway into the room he recognized us and paused for a
moment. He appeared stunned, and his eyes went wide.
   "You'reTsutomu!" he said, with surprise in his voice, and then
he looked at the reporter sitting next to me. ''And you're
   Both of us nodded.
                      "YOU'RE TSUTOMU1"                        309
   It had become clear to both Mitnick and me that this wasno
longer a game. I had thought of the chase and capture as sport,
but it was now apparent that it was quite real and had real con-
   Having spent several weeks on this man's trail, seeing the dam-
age he had caused, coming to learn that he was not only single-
minded in his invasion of other people's privacy and his pursuit
of their intellectual property, but also petty and vindictive, I
knew one thing for certain about Kevin Mitnick: he was in no
way the hero of a movie about some mistreated computer hacker
whose only crime was curiosity. There was nothing heroic about
reading other people's mail and stealing their software.
   He was led to the defendant's table and Judge Dixon came in.
The hearing was over in under ten minutes. A public defender
had not yet been appointed and so Mitnick sat at the defense
table alone with the u.s. marshal close behind him. I was curi-
ous to see if he'd continue his masquerade, but when asked his
name he identified himself as Kevin David Mitnick. The fight
had gone out of him, and he was clearly tired.
   As the judge read the charges-telecommunications fraud and
computer fraud, each carrying a maximum potential sentence of
fifteen years or more-it was clear that Kevin was beginning to
understand what was in store for him. This game had real penal-
ties. In a soft voice he said he wanted the court's permission to
contact his attorney in California. The judge noted that whatev-
er happened in his legal entanglements, the u.s. Court of the
Eastern District of North Carolina would "have its way" with
him first. The detention hearing was set for two days later on
Friday morning.
   The whole thing lasted less than ten minutes. After the judge
adjourned the court, Markoff made his way up to the railing that
separated the spectator gallery from the rest of the courtroom.
Julia and I followed him. Kevin rose and turned to face us.
   He straightened and addressed me. "Tsutomu, I respect your
skills," he said.
   I returned his gaze and just nodded. There didn't seem to be
much to say. In our contest he had clearly lost.


   Strangely, I felt neither good or bad about seeing him on his way
to jail, just vaguely unsatisfied. It wasn't an elegant solution-
not because I bought some people's claims that Mitnick was
someone innocently exploring cyberspace, without even the
white-collar criminal's profit motive, but because he seemed to be
a special case in so many ways. This was the sixth time he'd been
arrested. He certainly knew what the stakes were, and I hadn't
seen any evidence of a higher moral purpose to his activities or
even just innocent curiosity.
   The marshal started to lead him away and Markoff said,
"Kevin, I hope things go okay for you."
   Mitnick appeared not to have heard him at first, but then he
stopped for a second and turned back toward us. After giving a
slight nod of his head, he turned away and was led out of the
   The three of us walked back to the elevator. Kevin was in deep-
er trouble this time than ever before. He was a second-time
offender on at least two federal crime statutes, and he was a pro-
bation violator. More than half a dozen federal districts and sev-
eral states were waiting to file charges against him.
   None of us could figure out the psychology of his obsession.
Did he imagine himself the innocent voyeur that his friend Eric
Corley believed him to be? Or was he wrapped up in his own leg-
end, living out some Robert Redford-inspired vision of being a
last American hero on the run? Was he some new kind of cyber-
addict, as a federal judge decided back in 1988? I'd read some-
where that gamblers and check forgers exhibit similiar behavior:
even though they realize they will lose or be caught sooner or
later, they're trapped by an irresistible urge to keep going until
they fail. Maybe somewhere deep inside, Mitnick had been so
seduced by the game that he accepted the same fatal certainty
that sooner or later he would be caught. There was no way of
   Afterwards we drove back to the Sheraton. Markoff went off to
file his story for Thursday's New YOrk Times, and Julia and I spent
the rest of the day in the hotel room- much of it on the phone
with Andrew, officials at the Well and Netcom, and other systems
administrators around the Internet whose networks Mitnick
might have compromised. That evening, around 9 P.M., Markoff,
Julia, and I drove back to the Sprint switch and waited until
Murph and Joe could go out to dinner. By the time they got off
                      "YOU'RE TSUTOMU!"                        3I I
their shift it was late, and we drove around for a long time before
finding a place to eat.

   The next day I woke up to find that the real chaos had begun.
Kevin Mitnick's arrest was big news, and the media deluge was
happening in earnest. That morning I also discovered one last
message from the mysterious caller. It had been left on my office
voice mail at 7:23 P.M., San Diego time, the night before, in an
urgent voice:

   "Tsutomu, my friend. I just want to say ... want to reiterate it
is big joke. It is big, make fun of Kung Fu movies, has nothing to
do with computer hacking, Mitnick, nothing! You tell them, do
not send them to come and get me. No, do not fly out to come
and get me. I'm not worth it . . . just make fun of Kung Fu
movies. That is it. Thank you."

   The game really was over.


In July of 1995, Kevin Mitnick agreed to plead guilty to one
charge of cellular telephone fraud and, without a trial, was sen-
tenced to eight months in jail. As a result of his arrest in Raleigh,
Federal prosecutors had originally charged him with twenty-
three counts of telephone and computer fraud, but all but one of
the charges were dropped as part of his plea bargain.
   However, Mitnick's legal problems aren't over. He is now in jail
in Los Angeles awaiting new charges including federal probation
violations, tampering with the State of California Department of
Motor Vehicles computers, and computer and telephone fraud
charges that may be filed by more than a half dozen federal dis-
tricts. Unless he agrees to a new plea bargain, a trial is now sched-
uled for late November, and it is likely that he will spend more
time in prison.
   Today, as I look back on the events of the past year, I'm still
troubled. In the wake of his arrest in February, Mitnick's legend
continued to grow. For many months after his arrest, heated dis-
cussions about his actions took place on the Internet. Some peo-
ple continued to argue that because Kevin Mitnick never physi-
cally harmed anyone, what he was doing was innocuous.
   The fact is that this is also the case of a man who has had fif-
teen years and six arrests to figure out what is right and wrong.
In the late 1980s, a federal judge made a special effort to give
him a second chance.


   For me, Kevin Mitnick's real crime is that he violated the orig-
inal spirit of the hacker ethic. It's not okay to read other people's
mail, and to believe that software and other computer technolo-
gies should be freely shared is not the same as believing that it's
okay to steal them.

   The network of computers known as the Internet began as a
unique experiment in building a community of people who
shared a set of values about technology and the role computers
could play in shaping the world. That community was based
largely on a shared sense of trust. Today, the electronic walls
going up everywhere on the Net are the clearest proof of the loss
of that trust and community. It's a loss for all of us.

                                                  October 25, 1995