Researchers develop cheap facial recognition system to predict SSNs

Aug 2, 2011
3

Researchers at Carnegie Mellon University have devised a surprisingly simply method of taking a photo of someone's face and using that photo and off the shelf software they can predict the persons social security number from that photo and a few other details with surprising accuracy. It's hard to believe a photo of you and some software is all it takes to divine your social security number.

The head of the research team at CMU was Alessandro Acquisti. The team conducted several experiments and the first they data mined Facebook for photos of people with searchable profiles and then used those photos to create a database using PittPatt facial recognition software. The first phase of the experiment used a dating site and the team identified 15% of the people in photos harvested.

The second experiment used a cheap $35 webcam to take photos of CMU students and then gave the students an online survey to take. While the survey was taken, the software worked to find the students identity and 42% of the people out of the 93 participants were linked to a Facebook profile. Then using items for other experiments, the team was able to predict their SSN. The rub was 60% of the students were foreign and didn’t have a SSN to predict.

[via Forbes]


Must Read Bits & Bytes