The FBI is rapidly automating the process of identifying crime suspects by running photographs through a digital police lineup. Records released by the FBI in response to a Freedom of Information Act lawsuit by the Electronic Frontier Foundation (EFF) reveal plans to stick the mugs of almost one in six Americans into the Next Generation Identification program's facial recognition database by next year.
Combined with the feds' tolerance for a remarkably high false-positive rate, the chances of getting fingered for somebody else's misdeeds are getting pretty good.
According to Jennifer Lynch, senior staff attorney with the EFF, the FBI plans to have as many as 52 million photos in its database by 2015. The system is designed to be capable of conducting tens of thousands of searches every day. Those 52 million images will include a planned 4.3 million faces photographed for non-criminal purposes and another million drawn from ill-defined sources. Searches will be run against all records in the database, no matter how they were obtained.
"This is a problem because we do not know what rules govern these categories, where the data comes from, how the images are gathered, who has access to them, and whose privacy is impacted," warns EFF's Lynch.
Whatever the source of the images, there's no guarantee they'll be correctly matched to suspects. Last year, the Electronic Privacy Information Center extracted a separate set of documents from the FBI revealing that federal specifications on the Next Generation Identification system facial recognition software allow for tagging "an incorrect candidate a maximum of 20 percent of the time."