Similarity Attack on Cancelable Biometric Templates using Particle Swarm Optimization

Abstract

Biometrics models use behavioral and physiological characteristics for automatically recognizing the identity of individuals. Since the biometric information is immutable, it needs to be stored in a secure way. Cancelable Biometrics (CB) is one such technique that is used for the purpose of biometric template protection. CB schemes typically utilize a transformation function to convert the biometric features into their corresponding cancelable forms. Although the cancelable templates are irreversible, the distance between two feature templates gets preserved during the transformation process. This similarity preserving property of CB causes a Similarity-based Attack (SA), in which the pre-images of the transformed features can be reproduced. In this paper, the applicability of Particle Swarm Optimization (PSO) is analyzed for performing the similarity-based attack on cancelable biometric templates. During our investigation, we have found that not every metaheuristic can be efficiently used for attacking the CB scheme. Furthermore, the performance of the metaheuristic varies depending on the underlying CB scheme. Hence our study provides a detailed outline of the applicability of PSO for launching a SA on a generic CB scheme.