Facial recognition software is the law enforcement frontier. Rather than having to build a lineup, law enforcement can just run suspected criminals’ faces against the collected photos of criminals and non-criminals alike in hopes of a positive identification.
At this point, it’s still very touch-and-go. Technology hasn’t kept pace with law enforcement’s dreams of an accurate and speedy way of ID’ing suspects. As of 2008, the FBI was granting the contractor behind its system a 1-in-5 margin of error. Yes, a 20% chance of nabbing the wrong person was considered acceptable in a live system.
The technology continues to improve, but it still requires clear photos taken nearly head-on for best results. Despite these limitations, law enforcement agencies continue to take these systems live, almost always without putting together some sort of privacy/data disposal policy. But we’re all supposed to be fine with this because these agencies are using this tech to track down dangerous criminals and/or terrorists, right?
British cops used a new facial recognition system to snare a shoplifting suspect whom they say was automatically identified due to his resemblance to criminal relatives, The Register has learned.
And, apparently, shoplifters. Not only are British cops bragging up an expensive system’s ability to nab an extremely low-level criminal, it’s also playing up the fact that the system failed to pick the suspect out of the “lineup.” Instead, it just seized on the fact that the suspect resembled other criminals in its database. Not exactly comforting… at least not for citizens who may resemble suspected criminals and vice versa.
UK law enforcement, on the other hand, seems rather encouraged by the software’s inability to correctly pick out a shoplifting suspect from a digital lineup.
The Metropolitan Police Force is due to visit Leicestershire this week to scope out NeoFace, it is understood, while the Essex and Kent forces have already been to check out the system. French and Romanian officers have also been in contact to express an interest.
The saving grace of this imperfect system is that it can’t directly be used as evidence.* It can only guide a “line of inquiry.” The downside is that photos are retained for five years and the Leicestershire police seem very happy that the software is so good at detecting familial members, rather than the people they’re looking for.
*I’m sure the UK police are equally familiar with the concept of parallel construction.
On one hand, this isn’t entirely unlike the old photo books police use to identify suspects. On the other hand, NeoFace doesn’t just store photos of criminals. This is especially problematic in the UK, where CCTV wiring is the new kudzu. The Leicestershire Police has 90,000 photos in its database and that number should only be expected to expand rapidly, especially if coupled with NeoFace’s other offerings.
NeoFace Watch watches surveillance footage, constantly picking faces out of a crowd — and then storing those faces in a database, or matching them against a predefined watch list. NeoFace Smart ID is a smartphone and tablet app that allows for the real-time collection and identification of fingerprints, faces, voices, and other identifiable data at crime scenes.
We’re told these developments’ privacy implications are mitigated by the dangerous criminals they’ll be used to apprehend. And then it’s all undercut by law enforcement members excitedly talking about nabbing a shoplifter who sort of resembled two other people.