US Army soldiers learn about retina scanning in a biometrics class. Photo: Pvt. Luke Rollins
Like so many now-expansive government programs, the Department of Homeland Security's mass facial recognition initiative had benign origins. The NSA began as a code-breaking crew, remember, and the effort to comb American crowds for the faces of wanted criminals began as a military effort to weed out suicide bombers in Afghanistan.
What's now called the Biometrical Optical Surveillance System, or BOSS (note to bureaucrats: maybe stop giving creepy programs authoritarian acronyms), has been tested domestically and is inching towards deployment, according to The New York Times.
In that same Times dispatch, biometrics expert Anil Jain made an educated guess as to when that deployment might be. “I would say we’re at least five years off," he said, depending on what the feds plan to do with the system.
Five years away from the government being able to effectively scan crowds for faces that match those on, say, a terrorist watch list. Or the FBI's Most Wanted list. Or registered sex offenders. Or known recidivists. Or anyone with a criminal record. You're getting the point.
Once the technology to efficiently match an image of a face to one in a crowd is developed, we're instantly entered nebulous ethical territory. Paired with the recent revelations about the amount of personal data that federal agencies are currently accessing, the ramifications of the state deploying surveillance tech worry privacy advocates more than ever.
And rightfully so, though it shouldn't take the DHS's BOSS program to stir up concern, seeing as how facial recognition is being used by various government branches all over the place already.
The Department of State currently runs one of the largest facial recognition operations in the world. It uses a database of 75 million photos or so to cross-check visa applications. The DMV is using facial recognition technology to prevent identity theft. Staffers scan driver's license and ID applicants to see if there are already IDs awarded under that name.
Facial recognition is already in use around the globe for a wider range of activities, and has been for years.
The Department of Justice has been funding facial recognition tech for over a decade—though it hasn't really been paying off. Many remember the 2001 Super Bowl where the Tampa PD tried scanning the crowd, where 19 potential terror and criminal suspects were identified. Little progress in the way of policy solutions has been made since then, regardless of the controversy that ensued.
Mind you, that's just the US government. Facial recognition is already in use around the globe for a wider range of activities, and has been for years. It's used in Mexico to confirm voter identity. Australia uses it to check passports. Germany has created a centralized database for all the nation's police departments to access with facial rec tech. And some retail stores are already using the technology to profile certain kinds of shoppers.
Of course, a lot of that tech isn't ready for the prime time yet. Most recently and famously, perhaps, there was the infamous failure of the Department of Homeland Security's own BOSS software to locate the Tsarnaevs in the wake of the Boston bombings. Despite the hundreds of dollars that had been sunk into the BOSS program since its military days, and both Dzhokhar and his brother's images being contained in the database, it didn't successfully turn either of them up.
Ars Technica has a nice thorough look at why this was the case, and concludes that it's likely the inputs that are lacking—the tech isn't good enough to automatically match two basic images in such a vast database yet. The system needs not just frontal photos, but side shots and iris and retinal scans. Cameras can pick up iris movement, so it would help ID perpetrators (or whoever else) if that info was on file. Ars muses that it's conceivable that DMVs might one day record such information as part of a state identification or license requirement.
And once that part of the puzzle is in place, the actual image scanning will be a lot easier.
"The technology is currently at a state where these face recognition algorithms can be deployed in anything from cell phones to large multiserver search engines capable of searching over 100 million faces in just a few seconds with operational accuracy," Brian Martin, facial recognition software engineer with MorphoTrust USA, testified before Congress.
If Anil Jain is right, and the technologies align in just five years, it could drastically overhaul the surveillance state, for better and for worse. We already have precious few checks on the surveillance already conducted by authorities—the notion that as soon as 2018 they may be able to effortlessly pick our digital faces out of a crowd isn't likely to soothe anyone already uncomfortable with the NSA's prying eyes.