I've had my doubts about the FBI's facial recognition systems, but a new case has given a rare window into how the system will actually work — and it looks a lot less like science fiction than old-fashioned border control.
The story starts with a fugitive, Neil Stammer, who had been on the run since he was arrested for child sex abuse and kidnapping in 1999. According to police reports, Stammer speaks nearly a dozen languages, so when he fled the country using false papers, it left law enforcement with few clues as to where he might be. The case languished for 14 years until this January, when a new facial recognition system found a face that matched Stammer's. The Diplomatic Security Service had his face on file under the name Kevin Hodges, thanks to a recent visa application to the US Embassy in Nepal. The DSS contacted the FBI, and a few months later Stammer was in custody.
A visa application in Nepal tipped them off
It's a simple story, but it gives a sense of how the FBI is actually using facial recognition. As it turns out, it's less complicated than you might think. In Stammer's case, both of the photos were passport-style photos taken from close range, with the subject directly facing the cameras. That's the strike zone for modern facial recognition, so it's not surprising FBI technology was able to match the faces.
As a result, it's easy to overstate the importance of facial recognition to a case like this. The current tech can't search through crowds in a Boston Bombing scenario, and it's not clear it ever will be. Even systems like the one in Chicago end up blocked by bad angles and inconsistent lighting. If Stammer had grown a mustache or had plastic surgery, it's plausible that he could have slipped through the system entirely. The converse is also true: If the embassy staff in Nepal had been more diligent, they might have recognized Stammer without any tech at all. But the simple "Wanted" poster has always been a powerful tool — and in the age of computerized pattern matching, it may be more powerful than ever before.