Information-Theoretic Analysis of Iris Biometrics for Biometric Cryptography

This paper presents a rigorous information-theoretic analysis of iris biometrics with the aim to develop optimized biometric cryptosystems. By estimating local entropy and mutual information, we identify the iris regions that are most suitable for these purposes. Parameter optimization of the appropriate wavelet transform produces higher entropy and low mutual information in the transformation domain. This establishes an effective framework for the development of systems for the extraction of truly random sequences from iris biometrics, while not compromising its proven authentication features.