Face to Facebook
As DeepFace facial recognition rolls out on Facebook its still unclear exactly what the privacy implications may be.
Face to Facebook
It was only a year ago that Google Glass arrived with an app, NameTag, that, using the camera built into Google Glass, took photos, sent them (via the wearer's mobile phone) to a server, scanned numerous sites and came back with a match.
It offered a name, additional photos and social media profiles, not to mention a scan against names in the National Sex Offender Registry and other criminal databases. All of this, in theory, made for error-free dating, in the event your would-be date didn't mind being pre-assessed in this way.
Twelve months later, Google Glass is no longer with us. However, around the time that NameTag briefly gained notoriety, Facebook's DeepFace facial recognition technology was announced — and is now rolling out. This, arguably, is facial recognition technology on an altogether higher plane, involving 3D modelling and a neural network trained on a database of 4,000 faces across four million images.
The result is software able to identify a person almost as well as a real human being can (which is between 97 and 98 percent accuracy). The aim? Initially at least to suggest tags for a photo automatically and then — counter-intuitively perhaps — to help the subject of the photo to avoid having his or her picture shown. Facebook can, using DeepFace, identify its users in new photos (from hundreds of millions daily) as they are uploaded, tell those users, and give them the opportunity to do something about it. So if you are not happy about appearing in your friends' timelines you will be able to blur your face and retain your privacy.
Emma Carr, director of Big Brother Watch, an organisation whose stated aim is ‘exposing the true scale of the surveillance state by challenging the policies which threaten our privacy, our freedoms and our civil liberties,' says, “Tagging someone takes a matter of moments, so DeepFace seems like nothing other than an antagonistic waste of time.” She continues, “It is almost inevitable that Facebook won't be the only organisation that will want to use DeepFace. The police and security services around the world will be clamouring to have access to not only the technology, but to the millions of photos on Facebook's database.”
However, Sarb Sembhi, a director with Storm Guidance, and a leading light in ISACA, the not-for-profit IT security association, points out that, impressive as this technology seems, security services have already shown that there could be further to go. “The technology is static,” he says. “It's slightly behind the technology used in [some] videos where police and crime prevention are looking at certain behaviours and movements as well as the facial technology.”
Being able to identify mannerisms, height, behaviour, stance and gait would certainly be a major advance but what would it all be used for? For now Facebook will not be drawn on any other plans for its technology. In any case, DeepFace is not happening yet — at least in Europe, where it's not available for legal reasons (except perhaps if you're a European uploading a photo of an American friend). Elsewhere Facebook settings allow the feature to be disabled.
But the technology can only improve and be enhanced and the tagging concept seems a rather modest application of it. Could there be privacy implications?
Carr says: “With regulations failing to keep up with technology, the future is certainly unclear in terms of protecting internet users' privacy. Whilst companies fail to include privacy by design in all their new technology, this will only continue to be a problem.”
Sembhi adds that it's always wise to ask “What's the business model behind the service provider? Because that will tell you what the bottom line is. [Facebook is] trying to collect as much information about you as possible to serve you ads. If you know that, you know there are going to be privacy implications.”