Holistic-Based Feature Extraction And Error Correcting Code Biometric Template Protection Technique

Abstract

Biometric template protection is one of the most important issues in biometric authentication system. It is vital because once the biometric template is being attacked, the intruder could introduce him/her into the system without following the proper enrolment procedures. Previous related biometric template works have some limitations in terms of lower accuracy level and high correlations between templates. These two problems may lead to false accept attacks and crossmatching attacks on these templates. To mitigate these problems, this research aims to increase the accuracy level by reducing errors in biometric template and to reduce correlation between biometric templates by generating uncorrelated feature vectors to reduce error in biometric template in order to increase accuracy level and generate uncorrelated feature vectors. To reduce errors and to generate uncorrelated feature vectors, Error Correcting Code (ECC) based on combination of LDPC and RS and Holistic-based Feature Extraction based on LDA, PCA and ICA are designed respectively. The proof of concept is tested on Iris biometric templates using X number of samples from the Y benchmark image repository. The results showed that the proposed template protection technique is able to increase accuracy level by A% and reduce correlation between biometric templates to B%. Thus, this technique is a viable and practical template protection technique without degrading the iris recognition performance.