On Generating Cancelable Biometric Templates Using Visual Secret Sharing

Abstract

Cancelable Biometrics have been gaining popularity due to security and privacy concerns of an individual’s Biometric image(s). The main objective in Cancelable Biometrics is to generate templates which are non-invertible in nature and result from repetitive distortion of the Biometric image. In this paper, we propose a novel framework for generating Cancelable Biometric templates using Visual Secret Sharing. In the proposed scheme, n different shares are generated corresponding to one Biometric image with the help of $n-1$ other images called Cover images. The generated Secret Shares are stored in a distributed manner instead of the original Biometric image. For generating n shares, we propose three different methods (M1) One Biometric image and $n-1$ randomly chosen natural gray images (M2) One Biometric image with $n-1$ randomly permuted version of Biometric image (M3) Both Secret image and Cover images are randomly permuted version of the Biometric image. To show the efficacy of the proposed approach, we have used the publicly available IIT Delhi Iris database (version 1.0). The performance of these three approaches have been compared in terms of average Co-relation Coefficient, False Accept Rate (FAR), False Reject Rate (FRR) and Genuine Accept Rate (GAR), True Error Rate (TER) and True Success Rate (TSR). The experimental results show that M3 performs best among the proposed methods in terms of all the performance measures and in qualitative terms.