Facial Recognition Software Takes One Glance at You and Brings Up Your Facebook Profile

Brian Teutsch via Flickr

Worried about privacy on the Internet? It may be even worse than you thought — with rapidly improving face recognition technology, your automatically tagged Facebook pictures could help a stranger, or the authorities, quickly identify you on the street.

A simple system that compares Facebook pictures and webcam snapshots can make a positive match after less than three seconds, according to Carnegie Mellon University researchers. Alessandro Acquisti and colleagues presented their findings at the Black Hat computer security conference in Las Vegas.

“A stranger could know your last tweet just by looking at you,” Acquisti told CNET’s privacy blog.

The system was able to correlate Facebook profile pictures to webcam shots, and to otherwise anonymous photos on a dating website. The Facebook-webcam system identified about 31 percent of users, and only 10 percent of the dating site users, but the message was clear — anonymity is becoming harder and harder to maintain.

This can be problematic for several reasons, not the least of which is the damage that can be done by mistaken identity. Computer systems that put the wrong name to a face can cause headaches or worse. But in simpler terms, do average Facebook users really want random people to find out their e-mail addresses and phone numbers simply by looking at their faces?

Google engineers have discussed hypothetically using its own this technology for such a purpose, but say they have no plans to actually do it.

Acquisti, an information technology and public policy professor, made a database of about 25,000 photos taken from CMU students’ Facebook profiles, CNET explains. Acquisti had volunteers peer into a webcam, and facial recognition software connected their images to their Facebook profiles. The system made a successful identification for 31 percent of the students after less than three seconds. The team also compared about 278,000 Facebook pictures to 6,000 dating website profile pictures, in which the members used pseudonyms, and about one in 10 were identifiable, CNET says.

Then the CMU researchers also developed an iPhone app that works the same way, running a photo through facial recognition software and displaying that person’s name and information on the screen.

The system only works with front-facing photos, and it would need to be refined, but as technology improves it will only get easier, Acquisti said.