Securing templates in a face recognition system using Error-Correcting Output Code and chaos theory

Abstract

In biometric cryptosystems, biometric data is combined with cryptography algorithms to generate secure templates. In these systems, creating protected templates with both high discriminability and high security is a challenging issue. To address this issue, this paper proposes a new face cryptosystem based on binarization transformation, chaos feature permutation and fuzzy commitment scheme. To enhance discriminability, real-valued templates are converted into their binary versions using a new discriminant binarization transformation based on Error-Correcting Output Code. Then, the chaos feature permutation is used to increase the security and privacy of binary templates, and also to protect the fuzzy commitment scheme against cross-matching attacks. The proposed scheme is evaluated on three well-known face databases, i.e. CMU PIE, FEI, and Extended Yale B. Experimental results show that the proposed method improves discriminability, as well as privacy and security of the system, compared to the existing face template protection algorithms.