Binarized Revocable Biometrics in Face Verification

Abstract

A two-factor revocable authentication approach, combining user-specific pseudo-random bit sequence with biometrics is presented. Through the mixing process, a distinct binary code, coined as BioBit, is formed. This provides a protection layer against biometrics fabrication because the system can cancel the compromised and reacquire a new one via bits replacement. BioBit delivers lower error rate as compared to sole biometrics when the genuine token is used by the authorized user. There are two identity theft scenarios may be raised: 1. stolen-biometrics: an imposter possesses intercepted biometrics data to be considered genuine; 2.stolen-token: an imposter has access genuine token and used it to claim himself as the genuine user. BioBit scheme shows the impressive performance (EER=0.001% and EER=0.002% tested on ORL and FERET datasets) in case 1. For thestolen-token case, this approach attains EER=1.28% and EER=1.36% on ORL and FERET datasets.