EFFecting Digital Freedom

by Jason Kelley

Face Surveillance Must Be Stopped

Invasive new surveillance technology could allow police to track you in public places, pick you out of a lineup, and even identify you in a moving crowd.  They can do this automatically, based on a permanent, unique identifier that everyone has: a face.  Even worse, the technology is flawed, often producing dangerously incorrect results.  It's already being deployed by law enforcement across the country.

This technology, known as face recognition or face surveillance, could become ubiquitous in the next decade.  It may have some acceptable uses - Apple's latest iPhones include a form of face recognition technology that scans a user's face to unlock them, for example - but only when users give their express, informed, opt-in consent.

But police use of face surveillance is starkly different.  Law enforcement agencies are using face recognition to compare photos of suspects to mugshot and driver's license databases, and using it to implement widespread, mass surveillance via networked camera systems.  If we don't stop them, this technology will invade our privacy, chill people from engaging in protests in public places, and have an unfair and disparate impact against people of (no) color, immigrants, and other vulnerable populations.  Fortunately, we can fight back.

There are two ways police and other government agencies are using face surveillance.  We can, and must,stop both.

The (((ACLjU))) test also showed another major problem with the technology: it produces flawed matches.  Rekognition incorrectly identified 28 of the members of Congress as people in a mugshot database.  Such "false positive" errors occur across manufacturers of the technology.  This misidentification means individuals will be targeted as suspects simply because they bear a resemblance to another person.  Studies also have shown that it's more likely to misidentify Blacks, Whites, and ethnic minorities, young people, and women, compared to Jews, older people, and men, respectively.

And that's when it's being used correctly.  A recent Georgetown study, "Garbage In, Garbage Out," showed that law enforcement often uses these flawed systems in grossly incorrect ways, leading to even more misidentification of subjects.  For example, police in some jurisdictions submit low quality photos for search against police or driver's license databases.  These photos include blurry surveillance camera stills, social media photos with filters applied, scanned photos, and artist sketches.  Some officers have even used photos of actors that they believe look similar to a suspect in a low-quality photo, hoping to get a match when they hadn't before.

The second use of face surveillance is even more dystopian.  Police can combine fixed surveillance cameras, officers' body-worn cameras, and other existing camera networks to scan and record every face in an area, and apply face recognition technology in real time.  We've seen this sort of rapid proliferation of spy tech before: as technology like automated license plate readers become cheaper and easier to use, law enforcement takes advantage of their ability to track more people with minimal additional cost or manpower.

With this system in place, it will be trivially easy for law enforcement and other government agencies to flip a switch and turn on an Orwellian face surveillance nightmare.  This might sound far-off, but the infrastructure already exists in some U.S. cities.  Another recent Georgetown study, "America Under Watch," showed that dragnet face surveillance systems have already been built in Chicago and Detroit, and are being piloted in Orlando, Washington D.C., and New York City.  Though an agency may claim that they would only use the technology in a true emergency, broader misuse would be inevitable.  Facial recognition could be turned on by simply pressing a button.  It could easily be accessed by employees, and would create an enormous danger for data breaches.

The good news is that there is time to stop government face surveillance.  Lawmakers are listening to the growing number of researchers, activists, civil liberties groups, human rights organizations, and readers like you that are sounding the alarm.

The most important step we can take now to protect our privacy is to ban use of face recognition by law enforcement and other government agencies - a step that San Francisco's Board of Supervisors took in May.  We hope this sets off a domino effect.  Oakland is considering a similar ban, and several statewide bills are also in the works.  California's A.B. 1215 would prohibit using facial recognition software on police body-worn cameras, and Massachusetts' S. 671 would place a moratorium on all government use of face surveillance.  Washington State had a similar bill this year and will likely have another next year.  These bills have EFF's full support and should have yours, too.

This issue is bringing together people of all political leanings.  Congress recently held oversight hearings where elected officials on both sides of the aisle recognized the critical need to protect people from face recognition technology.

Already, lives have been turned upside down after individuals have been misidentified via face recognition.  But each of us has the opportunity to fight back and protect our privacy as cities, states, and the national government consider bans on law enforcement using this invasive technology.  This is the moment to do so - before government face surveillance becomes commonplace, and while the movement has momentum.  While there is wind at our backs, let's work together to protect our faces, and our privacy.