Worried about facial recognition technology? Here’s a simple-to-use de-identification solution

22 - April 2019


Yesterday, at the Netexplo Innovation Forum 2019 at UNESCO’s Paris HQ, I took part in the presentation of one of the companies I had scored highly in the selection process of 10 outstanding digital projects competing for the event’s Grand Prix, Israel’s D-ID, and I had the opportunity to meet one of its co-founders, Gil Perry. The company, set up about two years ago, works on the de-identification of photographs: its algorithms generate imperceptible modifications in images that prevent the faces shown in them from being recognized by facial recognition algorithms, which means they can be used in databases without risk of subsequent identification. As the company explains(link to video), this is important, because while we can easily change our passwords, it is very difficult to change our face.

Photography, to all intents and purposes, shows us as we are, but small alterations, for example, the distance between the eyes or other features, can trick biometric characteristics captured by facial recognition algorithms, meaning that the photo does not match our identity. The human eye would still recognize us, but the biometric parameters are erroneous, and therefore if it is included in an identification database, would not allow us to be identified.

Facebook, for example, already has the largest collection of identification photographs on the planet, and we are permanently surrounded by cameras capturing our image all the time, which means that if we can break the link between our identity and a photograph of us we gain a certain freedom, giving us back control over a technology that, in the words of Mark Zuckerberg, «has developed too quickly», preventing us from considering all its implications. The use of such technology can “poison” a database: if somebody starts uploading photographs tagged with their name to Facebook, the company will have several images with different biometric characteristics associated with that person, hindering their use for identification.

De-identifying a photograph is simple and quick and can easily done before uploading an image to a social network, or printing a photograph to be used for an identity document. Paradoxically, using a D-IDed photograph on our passport would be simple, but would not be recognized by a digital reader at an airport, although it would work at manual control points.

The company, based in Tel-Aviv, raised $4 million last year and already has several clients using its technology. D-ID says it wants to make the technology available to anyone who wants to use it, thus giving users greater control over their privacy.

I think there is huge potential for «de-identification» to help control rapidly developing technologies that trigger processes or possibilities we are still coming to terms with. Particularly interesting is the possibility of making it safer to share photographs on social networks: I don’t like the idea of warning young people of the dangers of uploading photographs, because it goes against something as natural as sharing a memory. Telling school children that under no circumstances should they upload their photos to social networks contributes to a paranoid, fearful society; instead, we can now use. a very simple process before uploading photograph, giving us a significant degree of freedom. If in addition, the company providing the technology wants to make it widely available, so much the better.