A newly released set of slides from the Snowden leaks reveals that the NSA is harvesting millions of facial images from the Web for use in facial recognition algorithms through a program called "Identity Intelligence." James Risen and Laura Poitras's NYT piece shows that the NSA is linking these facial images with other biometrics, identity data, and "behavioral" data including "travel, financial, behaviors, social network."
The NSA's goal -- in which it has been moderately successful -- is to match images from disparate databases, including databases of intercepted videoconferences (in February 2014, another Snowden publication revealed that NSA partner GCHQ had intercepted millions of Yahoo video chat stills), images captured by airports of fliers, and hacked national identity card databases from other countries. According to the article, the NSA is trying to hack the national ID card databases of "Pakistan, Saudi Arabia and Iran."
This news is likely to be rhetorically useful to campaigners against national ID cards in countries like the UK, where the issue has been hotly debated for years (my own Member of Parliament, Meg Hillier, was the architect of one such programme, and she, along with other advocates for national ID cards, dismissed fears of this sort of use as paranoid ravings).
The development of the's NSA facial recognition technology has been accompanied by a mounting imperative to hack into, or otherwise gain access to, other databases of facial images. For example, the NSA buys facial images from Google's Pittpatt division, while another program scours mass email interceptions for images that appear to be passport photos.
An interesting coda to the piece is that the NSA has developed the capability to infer location by comparing scenery in terrestrial photos to satellite images, which sounds like a pretty gnarly computer-vision problem.
Read the rest