A New and Practical Design of Cancellable Biometrics: Index-of-Max Hashing

Despite a variety of theoretical-sound techniques have been proposed to generate cancellable biometric templates, there is rarely practical solution that satisfies non-invertibility, revocability, non-linkability and performance simultaneously. In this paper, we propose a locality sensitive hashing inspired cancellable biometrics framework, namely "Index-of-Max" (IoM) hashing. Briefly, IoM hashing non-linearly transforms a realvalued feature vector into discrete index hashed code. We demonstrate two realizations from IoM hashing framework, namely Gaussian Random Projection based and Uniformly Random Permutation based hashing schemes. The discrete indices representation nature of IoM hashed codes enjoy several merits such as it empowers strong concealment to biometric information. This contributes to the solid ground of noninvertibility guarantee. IoM hashing is insensitive to the magnitude of features, hence are more robust against biometric features variation and its magnitude-independence trait makes the resultant hash codes scale-invariant, which is critical for matching and feature alignment. The experimental results demonstrate reasonable accuracy performance on benchmark FVC2002 and FVC2004 fingerprint databases. Moreover, the analyses justify its resilience to the existing and newly introduced security and privacy attacks as well as satisfy the revocability and unlinkability criteria of cancellable biometrics. Besides, the implementation of IoM hashing is also incredibly simple for practical applications.