A face template protection approach using chaos and GRP permutation

In biometric systems, template protection is a vital issue for preventing from identity theft. Fuzzy commitment scheme is a template cryptographic method providing secure templates by binding a uniform random key to a template. Fuzzy commitment scheme suffers from its lacks in privacy, in cancelability, and also in robustness against cross-matching attacks. To improve both the security and cancelability properties simultaneously, we present a novel template protection approach for face recognition systems based on fuzzy commitment scheme, permutated features, and chaos symmetric key. To permute feature vectors, we produce pseudo random numbers by using nonlinear chaos function to fill the control array of GRP permutation method. Even if the permutated template is compromised, it is possible to substitute it with a new permutated template by changing the initial conditions of chaos map. To evaluate our proposed approach, a series of experiments have been conducted on two well-known face databases ORL and Yale. We showed that the new feature permutation leads to more secure protected templates against decodability based on cross-matching attacks. The experimental results also showed that our proposed approach outperforms existing fuzzy commitment methods both in security and privacy aspects without influencing the accuracy. Copyright © 2016 John Wiley & Sons, Ltd.