A two-phased approach to reducing the false accept rate of spoofed iris codes

Abstract

In this paper, we demonstrate how to reduce the chance of a spoofed iris code being falsely accepted by an iris recognition system. We simulate the system attack by taking one of the registered iris codes from a subject set and mutating it by several different rates and presenting the resultant iris codes to our system. Our approach uses the k-nearest neighbors from a training set to the known spoof to establish a critical distance. Presented iris codes from our mutant set that have a Hamming Ratio when compared to the spoof that is less than the critical distance are rejected. Those that are falsely accepted are totaled to produce a Spoof False Accept Rate (SP-FAR). The second phase of our approach uses traditional iris code recognition to reduce the SP-FAR by rejecting those spoofs that were mutated to a degree such that they will not match any of the other iris codes in the training set.