Sent to us by a reader.
ID checks were a common response to the terrorist attacks of 9/11, but they’ll soon be obsolete. You won’t have to show your ID, because you’ll be identified automatically. A security camera will capture your face, and it’ll be matched with your name and a whole lot of other information besides. Welcome to the world of automatic facial recognition. Those who have access to databases of identified photos will have the power to identify us. Yes, it’ll enable some amazing personalized services; but it’ll also enable whole new levels of surveillance. The underlying technologies are being developed today, and there are currently no rules limiting their use.
Walk into a store, and the salesclerks will know your name. The store’s cameras and computers will have figured out your identity, and looked you up in both their store database and a commercial marketing database they’ve subscribed to. They’ll know your name, salary, interests, what sort of sales pitches you’re most vulnerable to, and how profitable a customer you are. Maybe they’ll have read a profile based on your tweets and know what sort of mood you’re in. Maybe they’ll know your political affiliation or sexual identity, both predictable by your social media activity. And they’re going to engage with you accordingly, perhaps by making sure you’re well taken care of or possibly by trying to make you so uncomfortable that you’ll leave.
Walk by a policeman, and she will know your name, address, criminal record, and with whom you routinely are seen. The potential for discrimination is enormous, especially in low-income communities where people are routinely harassed for things like unpaid parking tickets and other minor violations. And in a country where people are arrested for their political views, the use of this technology quickly turns into a nightmare scenario.
The critical technology here is computer face recognition. Traditionally it has been pretty poor, but it’s slowly improving. A computer is now as good as a person. Already Google’s algorithms can accurately match child and adult photos of the same person, and Facebook has an algorithm that works by recognizing hair style, body shape, and body language — and works even when it can’t see faces. And while we humans are pretty much as good at this as we’re ever going to get, computers will continue to improve. Over the next years, they’ll continue to get more accurate, making better matches using even worse photos.
Matching photos with names also requires a database of identified photos, and we have plenty of those too. Driver’s license databases are a gold mine: all shot face forward, in good focus and even light, with accurate identity information attached to each photo. The enormous photo collections of social media and photo archiving sites are another. They contain photos of us from all sorts of angles and in all sorts of lighting conditions, and we helpfully do the identifying step for the companies by tagging ourselves and our friends. Maybe this data will appear on handheld screens. Maybe it’ll be automatically displayed on computer-enhanced glasses. Imagine salesclerks — or politicians — being able to scan a room and instantly see wealthy customers highlighted in green, or policemen seeing people with criminal records highlighted in red.
Science fiction writers have been exploring this future in both books and movies for decades. Ads followed people from billboard to billboard in the movie Minority Report. In John Scalzi’s recent novel Lock In, characters scan each other like the salesclerks I described above.
This is no longer fiction. High-tech billboards can target ads based on the gender of who’s standing in front of them. In 2011, researchers at Carnegie Mellon pointed a camera at a public area on campus and were able to match live video footage with a public database of tagged photos in real time. Already government and commercial authorities have set up facial recognition systems to identify and monitor people at sporting events, music festivals, and even churches. The Dubai police are working on integrating facial recognition into Google Glass, and more US local police forces are using the technology.
This is a topic I’ve researched quite a bit. These systems are disturbing and should not be overlooked, but they don’t seem to be as robust as Schneier claims. (In fairness, he says they’re “being developed.”) Frontal images in good lighting work pretty well, but non-ideal conditions and a large number of “candidates” in the database will degrade the accuracy. Mirrored sunglasses are also reported to cause problems.
I suspect that if efforts to develop facial recognition systems based purely on camera images were so successful, there wouldn’t have been efforts by the Department of Fatherland Security to develop the system known as “BOSS,” which uses much more sophisticated hardware. BOSS was in the news a couple of years ago. Even with dual cameras and some kind of laser or radar distance-finder, its accuracy was less than impressive.
I’m not familiar with the Facebook system, but it’s possible that its accuracy is greatly enhanced simply by having relatively few candidates to choose from when making an identification; i.e., it only picks someone out from your pool of “friends.” Tagging people in photos wouldn’t even be necessary if the computers could so easily identify everyone.
Regardless of FR’s current state of development, a curse on all those who have been working on it! I don’t look forward to the day when everyone who wants anonymity has to walk around in a full hood and mask.
Yes, BMF, I’ve heard similar reports of the technology’s unreliability, but they keep throwing it in the papers to support the “all seeing eye” propaganda that’s so effective at silencing dissent.
Yeah, it’s possible that some of the hype is a psy-op. I once saw an article claiming a system had been developed that could reliably pick people out of a crowd. It was dated 1997. Yet there have been numerous high-profile failures of FR in recent years: the London Riots, the Boston Bombers, etc. People still get away with robbing jewelry stores in Midtown Manhattan, easily one of the most heavily surveilled places on earth (including prisons). Chicago, another heavily surveilled city, has one of the lowest murder clearance rates (something like 25%, if memory serves).
Whether or not they get these systems working seamlessly, anyone who’s going to do something “unacceptable” in public knows to wear a full-head mask, indistinct clothing, etc. Bank robbers were doing this long before computers were ever invented. If the SHTF in a big way, anonymity might be a lot less important.
Those LED headlights clipped onto a hat also cause problems for facial recognition systems.
-Just wait a few years, until the ‘telemarketers’ start harassing you on the street.
(that’s just my own theory, for the record…and now that I wrote that line the spies will be messaging their bosses excitedly)