All Iris Filters are Not Created Equal

Previous research has assumed that all parts of an iris code are equally valuable. Alternatively, some researchers claim that parts of the iris are more valuable, but they still use the same portions of the iris for all subjects. Our research, presented originally, is the first and only work to show experimentally that some bits in the iris code are less reliable than others, and these patterns of fragile bits vary on a subject by subject basis. This paper extends the work presented, answering questions such as "how does a different filter affect the patterns of fragile bits?" and "what causes outliers in these distributions?" We present a modification to our iris biometric system that masks fragile bits and significantly shifts the match distribution away from the non-match distribution.