Attacks on Privacy-Preserving Biometric Authentication

Abstract. Biometric authentication based on facial image, fingerprint, palm print, iris, retina, or veins are becoming increasingly popular. How-ever, compromised biometric templates, indeed, may lead to serious threats to identity and their inherent irrevocability makes this risk even more se-rious. Because of such serious privacy implications the need for privacy-preserving biometric authentication protocols is of utmost importance. Recently, Yasuda et al. [1, 2] proposed two efficient privacy-preserving biometric authentication using packed homomorphic encryption based on ideal lattices and on ring learning with error. We review these proto-cols and analyse their security against malicious internal adversaries. Yasuda et al. [1, 2] have proposed two packed homomorphic encryption schemes based, respectively, on ideal lattices and on ring-LWE (ring-learning-with-errors). Let vE1p¨q be the type 1 packed encryption, and vE2p¨q the type 2 packed encryp-tion. Let A and B be bitstrings of length N. Then, ctH “ CvE1pAq`C 1vE2pBq´ 2vE1pAqvE2pBq corresponds to an encryption of the Hamming distance betwee