OBRE: Optimal Biometric Recognition uner Encryption
Funded by: NWO
Period: Jun, 2019 - May, 2020
Biometric data, such as fingerprints, are personal data. Their storage, especially in large, central databases, entails privacy risks, against which the fingerprints must be protected. GenKey wants to improve its current technology to protect stored fingerprints. Biometric recognition under homomorphic encryption is an excellent candidate for this because it offers the possibility of performing biometric recognition with encrypted fingerprints. A basic method for biometric recognition under homomorphic encryption has been devised at the University of Twente (UT), based on quantized likelihood ratios, Encrypted Quantized Likelihood Ratio comparisons (EQLR) and a comparison protocol, which differs from existing methods based on for homomorphic encryption through much more reliable and faster recognition. This basic method assumes a server on which the encrypted biometric data, the template, is stored and a client with a fingerprint sensor. Typical for the approach is that both client and server learn nothing about the stored template during the recognition, that the server does not learn anything about the biometric data offered and that only the client learns the outcome of the recognition in the form of an accept or reject. Initial experiments on fingerprint data from GenKey prove very promising. The aim of this project is therefore to make this method suitable for application by GenKey in their fingerprint recognition systems.