Fragmentation for Iris Template Protection

Existing solutions for protecting biometric templates are typically based on encryption or feature transformations, which may increase system’s complexity or degrade its accuracy. In this paper, we put forward the idea of departing from encryption and protecting biometrics through fragmentation. By partitioning templates into unlinkable fragments, fragmentation provides protection since the limited information contained in each fragment is not sufficient for correct identification and hence cannot be exploited to impersonate legitimate subjects. In storage, fragments are mapped to fragtemplates on which the matching process is then performed. In each fragtemplate, genuine fragments are confused with synthetic ones and are indistinguishable from them. We design a fragmentation approach for iris templates and experimentally evaluate its accuracy and protection guarantees. Our experimental results confirm that fragmentation, when properly designed, can effectively provide protection while preserving identification accuracy.