LONDON – Biometrics researchers are working on a privacy solution for facial data that would see smartphone user images encrypted into two separate encrypted files which are then also “hidden” in new, unrelated faces and stored seperately.Using a technique known as visual cryptography, two facial data templates are created from a single face. These templates are then “hidden” in an unrelated face – for example a celebrity mugshot, with one kept on a device and another in the cloud.Arun Ross, Associate Professor of Computer Science & Engineering at Michigan State University, who gave a talk on the process at Biometrics 2015 conference in London on Tuesday, said the process could have significant benefits in terms of biometric data privacy – particularly for mobile identity.Addressing the issue whereby hacked mobile devices could reveal facial data stored on them for biometric authentication, the technique could eliminate the risk of reverse engineering from templates or even from secure elements.While visual cryptography has been proposed as a solution to facial privacy before, this method's use of “host face images” adds an extra layer of security and privacy – the use of these faces means that social networks cannot mine the data for demographics.Ross told Planet Biometrics that the initial application envisioned was that facial data could be separated into two encrypted images and then stored by two agencies – effectively giving one agency oversight over the other although each “retains” a small part of a digital identity.Alternatively, mobile identity providers could store half of a biometric face on a device, and another part on a cloud.The process is outlined in this slide from the presentation:The two public face images can then be stored independently in different on-line servers so that the identity of the private image is not revealed to either server, and only when the two are reunited and decrypted can the face data be replicated.Ross said that there are plans to develop a technological solution using the proposed method.
'Hidden faces’ proposed as a biometric privacy solution
