Faces

I was reading about facial recognition technology recently in an article published by Cargegie Mellon University.  I didn't realize how far it had come.  It's a little spooky.  

In the olden days, it took a person to recognize a face.  Now it's done through augmented reality, which means computer-generated input plays a big role.   It's currently possible, with greater and greater degrees of accuracy, to mine publicly available online social network data with face recognition technology to produce automated, real-time profiles of people. 

Suppose you have your picture taken and you don't even know it because you're part of a crowd.  Maybe you're out shopping.  Maybe you're in somebody's Flickr photograph, have a profile on match.com, or were captured on a webcam.

Now suppose you have a Facebook page with a photograph of yourself.  Software finds a match, and they now have the name you used on Facebook, presumably your real name, along with everything else you've revealed about yourself.   Facebook is a particularly good way to make a match because their privacy policy does not allow privacy settings on your name and profile photograph.

If you're not on Facebook, maybe you're LinkedIn, or part of a government or corporate database.  Anything with a photograph tied to your name can be used to identify you.  It's even possible to take an anonymous photo from a dating website, run it through facial recognition technology, and find a match to a publicly identified person on Facebook. 

Some of the data is easy to find and directly tied to you, like your Facebook friends and interests.  Many websites keep information about you and they also sell the information to third parties.  

Other sensitive information, non-publicly available, such as credit scores, SSNs, political and sexual orientation can be correctly inferred as part of what is known as data accretion (the gradual accumulation of individual parts until a whole picture is built.)

Facial recognition, just one piece of data accretion, is used by Google's Picassa, Apple's iPhoto, and Microsoft's Kinect.  All those photographs that are stored on the cloud are fair game for mining.

 

Face recognition is hampered when your eyes and/or nose bridge are covered. And of course, the less information you reveal about yourself online, the less can be known.

For more information, watch this 15-minute TED talk, Why Privacy Matters, by Carnegie Mellon's Professor Alessandro Acquisti.