The US Federal Bureau of Investigation has begun rolling out its new $1 billion biometric Next Generation Identification (NGI) system. In essence, NGI is a nationwide database of mugshots, iris scans, DNA records, voice samples, and other biometrics, that will help the FBI identify and catch ‘criminals’ — but it is how this biometric data is captured, through a nationwide network of cameras and photo databases, that is raising the eyebrows of privacy advocates.
Until now, the FBI relied on IAFIS, a national fingerprint database that has long been due an overhaul. Over the last few months, the FBI has been pilot testing a facial recognition system — and soon, detectives will also be able to search the system for other biometrics such as DNA records and iris scans. In theory, this should result in much faster positive identifications of criminals and fewer unsolved cases.
According to New Scientist, facial recognition systems have reached the point where they can match a single face from a pool of 1.6 million mugshots/passport photos with 92% accuracy, in under 1.2 seconds [PDF]. In the case of automated, biometric border controls where your face and corresponding mugshot are well lit, the accuracy approaches 100%. Likewise, where DNA or iris records exist, it’s a very expedient way of accurately identifying suspects.
So far, so good — catching criminals faster and making less false arrests must be a good thing, right? Well, yes, but there are some important caveats that we must bear in mind. For a start, the pilot study has only used mugshots and driving license photos of known criminals — but the FBI hasn’t guaranteed that this will always be the case. There may come a time when the NGI is filled with as many photos as possible, from as many sources as possible, of as many people as possible — criminal or otherwise. This might be as overt as parsing CCTV footage and collating every single face into a database; or maybe you’re just unlucky and your face ends up in the system because you’re in the background of a photo starring a known criminal.