Lattice-Based Secure Biometric Authentication for Hamming Distance

Abstract

Biometric authentication is a protocol which verifies a users authority by comparing her biometric with the pre-enrolled biometric template stored in the server. Biometric authentication is convenient and reliable; however, it also brings privacy issues since biometric information is irrevocable when exposed.

In this paper, we propose a new user-centric secure biometric authentication protocol for Hamming distance. The biometric data is always encrypted so that the verification server learns nothing about biometric information beyond the Hamming distance between enrolled and queried templates. To achieve this, we construct a single-key function-hiding inner product functional encryption for binary strings whose security is based on a variant of the Learning with Errors problem. Our protocol consists of a single round, and is almost optimal in the sense that its time and space complexity grow quasi-linearly with the size of biometric templates. On implementation with concrete parameters, for binary strings of size ranging from 579 to 18,229 bytes (according to NIST IREX IX report), our scheme outperforms previous work from the literature.