Off-the-Shelf Facial Recognition
You don't need the power of a government or an Internet behemoth to make facial recognition work for you. At this year's Black Hat security conference (held in Las Vegas in August), a team of researchers from Carnegie Mellon University demonstrated how much they could accomplish with existing off-the-shelf technology.
The team took photos of people's faces and pushed those images through an off-the-shelf facial recognition program called PittPatt (which Google recently acquired). In the demonstration, in less than 3 seconds, the program compared the CMU researchers' photos to images publicly available on Facebook and returned 10 possible matches, along with the names of the matches. The process proved to be accurate more than 30 percent of the time.
The team then used information gleaned from Facebook profiles to guess the birth dates or birthplaces of the people that the software had accurately identified. With that information, they predicted the first five digits of each person's Social Security number and were accurate about 27 percent of the time.
"The bigger picture here was to show that we're getting closer to a world where online and offline data blend seamlessly, where you can start with an anonymous face in the street and you can end up identifying something extremely sensitive about the person by combining these different technologies," says the leader of the team, Carnegie Mellon assistant professor Alessandro Acquisti.
It's Not Just Big Brother--Watch Out for Little Brother
While the demonstration by Acquisti's crew may make anyone who cares about privacy queasy, the concepts used in the demo aren't ready for "Little Brother" yet. "If you asked me if I could go out into the streets of New York and identify anyone and everyone, the answer is no," Acquisti says.
That's because the off-the-shelf system that the researchers used won't scale to a task of that magnitude. "If you wanted to identify anyone in the street of a large city, you'd need a database of hundreds of millions of people, and--given the computational power available now--it's still not possible to do these face match-ups in real time," Acquisti explains.
Still, because so much facial information is available online at places like Facebook and Flickr, preventing that information from being used to intrude on individual privacy is almost impossible, according to Harry Lewis, a computer science professor at Harvard University. Lewis told PCWorld: "A private individual in a public--but what was previously thought of as anonymous--place is no longer going to find themselves anonymous."
People are quick to express concern about technologies like facial recognition in the hands of Big Brother, Lewis acknowledges. "But let's not get so worried about Big Brother that we forget about the fact that Little Brother is going to able to do exactly the same thing," he says.
Lewis also points out that, in principle, Big Brother can be controlled through regulation and legislation, but "we can't regulate what Little Brother does about public information, unless we want to surrender our civil rights of freedom of speech."
Closed-Circuit Cameras: A Precedent
Some people argue, however, that anonymity began eroding long before facial recognition appeared on the scene. The proliferation of closed-circuit television cameras is an example of that trend. "The Big Brother thing is just technology catching up to what's always been there," says George Brostoff, founder and CEO of Sensible Vision in Covert, Michigan.
Sensible Vision makes facial recognition software designed for authenticating a person's identity. When users install Sensible Vision's software--Fast Access--on their computer and then sit in front of that PC, the software recognizes their face and logs them in automatically. If a user leaves the computer, the software detects his or her absence and prevents anyone else from using the unit. The company sells both personal and enterprise versions of the software.
In the long run, many problems involving potentially invasive technologies such as facial recognition simply work themselves out, according to Stewart Hefferman, CEO of OmniPerception, of Guilford in the UK, which makes object and facial recognition software.
"There are ways, through technology and legislation, of making sure that people's privacy is protected while deriving the benefits of a technology," Hefferman says.
Staff Editor David Daw of PCWorld contributed to this story.