Protected Honey Face Templates

Abstract

Most existing biometric template protection schemes (BTPS) do not provide as strong security as cryptographic tools; and furthermore, they are rarely able to detect during a verification process whether a probe template has been leaked from the database or not (i.e., being used by an imposter or a genuine user). By using the “honeywords” idea, which was proposed to detect the cracking of hashed password database, we show in this paper how to enable the detectability of biometric template database leakage. We add an extra layer of protection since biometric features cannot be renewed. The biometric system design implies that protection mechanisms must satisfy the irreversibility property and to this respect we apply different correlation tests to show the non-distinguishability between genuine and honey templates. In this paper we implement the idea of a honey template protection scheme on faces and evaluate the security and accuracy performance.