Homeland Security and the FBI are developing facial identification software. The author, Ginger McCall, suggests two policies to restrict the use of such technologies.
First, facial-recognition databases should be populated only with images of known terrorists and convicted felons. Driver’s license photos and other images of “ordinary” people should never be included in a facial-recognition database without the knowledge and consent of the public.
Second, access to databases should be limited and monitored. Officers should be given access only after a court grants a warrant. The access should be tracked and audited. The authorities should have to publicly report what databases are being mined and provide aggregate numbers on how often they are used.
I approve of the second policy 100%, but I wonder whether having the faces of ordinary people on file would tend to make the database more effective in identifying criminals. In some areas, having tons of data give systems greater power; when Netflix knows that I like movies A, B, and C, and that several hundreds of people who like A, B, and C also like movie D, it's a good guess that I'd like D, too. Is there an analog to this in facial recognition?
Posts
