Secure IRIS Verification

Abstract

In this paper, we present a novel secure iris verification system, where a transformed version of the iris template instead of the plain reference is stored for protecting the sensitive biometric data. An error correcting code (ECC) technique is adopted to perform the comparison in the transformed domain. A two-segment method is proposed to execute the feature verification, where a Bose-Chaudhuri-Hochquenghem (BCH) code of a random bit-stream is introduced to eliminate the considerable differences between the features extracted from different scans of irises. A reliable bits selection process during the iris feature generation stage reduces the system error rate from 6.0% to 0.8%. The appropriate size of the set of reliable bits is determined by investigating the best match between the associated error correct cutting edge and the actual verification accuracy.