Microphones, Laptops, and Supertaps

by Dark Spectrum

PC microphones are everywhere.

They're in the home, the workplace, and in schools.  You often see omni-directional mics like the Labtec Verse 303 or AM-232 mounted high up on computer monitors.

You're careful what you say near them since you know how good their room pickup is and how easy it is to capture the audio stream from a PC mic.

After reading this article, you'll watch what you say near any mic.

The PC might have a benign or even trustworthy owner, but how can you be certain it hasn't been compromised by a third-party eavesdropper?

If you think about it, the idea of a hijacked mic is frightening.  It's much more effective than a wiretap - it can be set up from thousands of miles away and uses existing, innocuous-looking equipment to create a 24/7 monitor on an entire room or office cube.  Call it a "supertap."

When you see a lab, office, or school room full of PCs with omni mics, it's time to think back to Robert Heinlein's classic The Moon is a Harsh Mistress.  The only difference is that the PC mics are loosely connected via a network of systems rather than directly to a single computer.

What could anyone possibly do with such an overwhelming stream of information?  Lots of things: simple old Voice-Operated Transmission (VOX) or the newer Voice-Activity Detection (VAD) techniques can reduce the bandwidth a lot.  Specific speakers or topics can be picked out via speaker recognition and speech recognition technologies.  Simple correlation-based methods can track a specific individual through a field of microphones.

O.K., so much for omni mics.  But what about the others?  (And there are lots of them.)

Directional monitor-mount mics like the Labtec Verse 313/AM-240 or the directional desktop boom mics?  Close-talking mics used in those PC headsets you see lying on desks or hanging from cube partitions?  Don't forget that almost every laptop has a tiny built-in mic which is exposed when the laptop is open.  But what if the laptop is closed and buried in a docking station, or left disconnected and lifeless on a conference room table?

The chilling truth is that any of the above configurations makes a perfectly good bug for the PCs immediate vicinity, and some of them are effective enough to form the basis of a supertap.  It doesn't take any rocket science, either.  All that's necessary is to use 16-bit audio and adjust all recording gains to their maximum values.

The only black magic is in the dynamic range provided by 16-bit audio.  Most PC audio systems lose three or four bits to noise, but that still leaves you with at least 12 usable bits.

You can record an almost-inaudible -48 dB signal (0.4 percent of full scale), boost it by 256 to normalize it, and still have four bits or 24 dB of signal available.  The high gain will create highly amplified noise, and the four-bit speech won't sound good, but it will certainly be intelligible.

Don't believe me?  Then why not just try it to see what you pick up.  It's easy.

Use the Recording Control panel (SNDVOL32.EXE) to make sure the mic is selected, and to set its gain to max.  If you have a laptop then it might have a dual-purpose line-in/mic jack and in that case you should click on the Advanced button to verify that the microphone boost is enabled.

Use your favorite audio editor for recording.  If you don't have one, then you could use the basic Windows recorder (SNDREC32.EXE) but two much better choices are Cool Edit (www.syntrillium.com) and GoldWave (www.goldwave.com).

Whatever editor/recorder you're using, configure it to 16-bit mono audio in linear PCM format.  Your system might be able to get good recordings at 8 kHz but for now just play it safe and set the sample rate to 11.025 kHz or 16 kHz.

You need good audio output to hear the results.

Headphones are best, but external speakers are also good.  You will probably have to boost the output level.  That can be done via your headphone/speaker volume controls and system playback gain controls (SNDVOL32.EXE again) but you'll get less distortion if instead you use Cool Edit or GoldWave to normalize the audio before playing it back.

There are two microphone configurations that are particularly challenging: high-quality PC headsets and docked laptops.

Cheap headsets are no problem.  They pick up any sound, from any angle, in any position.  High-quality headsets with close-talking mics don't.

For example, the Andrea Electronics NC-65 stereo gamers headset with anti-noise features seems to live up to its claims.  Even so, it records ordinary speech five feet away as -48 dB and as already calculated that's all it takes.  The background noise is steady (wide-sense stationary to you DSP types) which means it's easy to develop a custom speech detector for it.  Chalk up any PC headset as... supertap-capable.

For a long-term test you'll need to record to disk and use a speech detector.  Those features are found in utilities developed by scanner/ham radio hobbyists, examples being Scanner Recorder (www.davee.com/scanrec), Vox Recorder (ninopo.freeweb.supereva.it/radio/VoxRecorder), and RecAll (www.sagebrush.com/recall.htm).

Docked laptops don't work as well.

There are two reasons for that.  First of all, high frequencies are attenuated by the narrow passages the sound has to pass through to reach the mic.  That makes consonants harder to understand, masks some of the cues people use to recognize speakers, and reduces faraway speech to meaningless mumbles.

The second problem is that the mic might have lots of noisy neighbors in there: fans and disk drives.  Fans produce continuous noise due to air flow.  Disks emit transient clicks that are hard to filter out since they aren't a steady noise; if you're experimenting with a built-in laptop mic then don't log the audio to disk.

For a worst-case scenario consider the (aging) Dell J650: its docking station is fully enclosed on three sides and the mic is centered above the keyboard far away from any open air, but it can still pick up speech from the immediate vicinity.

Newer Dell laptops use open-frame docking stations with the mic on the right side of the keyboard so it's much closer to free air and therefore produces better recordings.

I'll close off by explaining the "disconnected and lifeless laptop."

Modern laptops have power management features which allow you to configure how they behave when the case is shut.  It's sometimes possible to configure them to simply keep on running when closed up.

That still leaves those blinking LEDs, but any doofus with a screwdriver and wire cutters can disable them.  What's left is a high-capacity, highly configurable data logger.

It isn't likely to be hijacked by a third party, but it's still worth mentioning as a mic to be wary of.