Generating Cancelable Palmprint Templates Using Local Binary Pattern and Random Projection

Abstract

A compromised biometric is compromised forever and the user is no more secure in other database also. This challenge can be addressed by Cancelable Biometric. A Cancelable Biometric is distorted version of a original biometric, that can be canceled and reissued like a password, and also is unique for every application. Cancelable biometric for original biometric has been seriously understudied problem. This paper presents a novel Cancelable coding scheme based on Local Binary Pattern (LBP) and Random Projection, where the biometric features are distorted in a revocable but irreversible manner by first transforming the raw biometric data into a fixed-length feature vector and then projecting it onto randomly selected subspace using set of random numbers. Three methods has been presented and experimented thoroughly. The Cancelable palmprint with Local Ternary Pattern (LTP) and two random number set outperform the other two methods. The proposed scheme has been verified under the best case and worse case scenarios (normal and stolen token scenario) on the PolyU Database.