Iris Cancellable Template Generation Based on Indexing-First-One Hashing

摘要

Iris recognition system has demonstrated its strong capability in performing personal verification and identification with promising recognition accuracy. However, the conventional iris recognition system stores the unprotected iris templates in a database, which is potentially being compromised. Even though biometric template protection provides a feasible solution to secure biometric template, a trade-off between security and recognition accuracy is incurred. That is, the higher security level always trades with poor recognition accuracy and vice versa. In this paper, a new iris template protection scheme is proposed, namely "Indexing-First-One" (IFO) hashing. IFO hashing transforms the binary feature into index value with Jacaard distance preservation. The resultant template offers a good indication of inheriting similarity from the IrisCode and strong concealment of IrisCode against inversion attack as well as other major security and privacy attacks. Experiments on CASIA-v3 data set substantiate that the proposed scheme can achieve as low as 0.54 % equal error rate and well preservation of recognition performance before and after IFO hashing.