Nikon recently upset quite a lot of people with their new facial recognition technology – an anti blink system that automatically warns you if you have taken a picture of someone when their eyes were shut. The problem? It can’t recognise Chinese or Japanese faces. Users in Oriental countries or users taking photos of people with naturally almond shaped eyes found that the face recognition tech repeatedly prevented them from taking a photo – displaying the rather inappropriate message (in the circumstances anyway) “Oops! Someone blinked!”
In the US, various State departments have been having similar problems with their use of face recognition. Here we are talking about genuine facial recognition, which is supposed to be capable of identifying a single human being by – you guessed it – recognising their facial features.
A man called John Gass was summoned to court in Massachusetts to have his driving licence revoked for a series of unpaid parking fines – none of which he had received. The real recipient looked a bit like John Gass: a computer scanning millions of database entries and using facial recognition to match potential fraudsters to their crimes had erroneously identified Mr. Gass as the culprit.
Unsurprisingly, the law came down heavily on the side of the computer – resulting in a long and frustrating series of attempts to have the problem rectified. It was, eventually, but the problems experienced by both Mr. Gass and the law enforcement authorities (who clearly don’t want to waste time and money trying to prosecute innocent people for crimes they didn’t commit)are illustrative of the dangers of relying on facial recognition technology – and of the sometimes damaging ways in which tech errors are pointed out.
It doesn’t take a massive stretch of the imagination to come up with several ways in which facial recognition software can be foiled. Putting on weight; growing a beard; shutting your eyes. And , of course, accidentally happening to look quite like someone else – a trait shared by a lot of people. You only need think about the amount of times you have seen someone who looks like someone you know for that truth to sink in.
Ultimately there’s a question here about the reliability and value, not to mention the proper uses, of facial recognition technology. Even in its most sophisticated form, the programming is instructing computers to look only at signifiers – eye spacing, cheekbones, mouth width and so on – all of which can be easily transformed by people who have the most interest in fooling the detectors. As a form of biometrics it seems more useful as a backup to other sets of data rather than the be all and end all datum used to prosecute someone for a crime.
And here’s something that you might not like at all about facial recognition technology: social networking sites, which routinely store millions of images of your face (in your profile pictures, in the pictures you upload and in the pictures of you that other people tag) have the potential to sell your image for the purposes of facial recognition. There is as yet no evidence that Facebook has actually done anything of the sort – but researchers at Carnegie Mellon University in the States recently did a dummy run of a facial recognition scheme using Facebook profile pictures (by consent) and scored a hit rate of more than 3 per cent.
Given the limited time and basic techniques used in this trial it seems likely that a more concerted effort could have Facebook profile pictures working in conjunction with the legions of hidden cameras we know watch us in most public spaces – raising issues about privacy that I almost hesitate to think about. For now it’s enough to know that Facebook specifically states, in its privacy statement, that no privacy settings apply to either your profile picture or your name. In other words, the biggest website on earth can do whatever it likes with your face.
At the moment facial recognition software only really works with a frontal shot, like a passport or driving licence photo. The next generation of tech, though, is working on the side profile and oblique angles. Watch this space and look out for your face!
About the Author:
The above article is composed and edited by Rosette Summers. She is associated with many Technologies communities as their freelance writer and adviser. In her free time she writes articles related to mobile applications, social media, wireless broadband etc.