Cancelable Face Biometric Verfication Algorithm Built on GoogLeNet and Characteristic Arbitrary Projection

Abstract

Biometric pattern information securing is necessary for avoiding individual confidentiality and identity loss. Arbitrary projection built on cancelable face biometrics is an effective and efficient approach to accomplish biometric pattern securing. Nevertheless, standard arbitrary projection-built cancelable pattern design can be easily attacked by attackers. To resolve this problem, in this paper, a characteristic arbitrary projection built on a cancelable face biometric algorithm is proposed, in which the projection arrays are created from one fundamental array in integration with face biometric information. The created projection arrays are deleted after utilization which makes it impossible for the hacker to attack information. The proposed algorithm is tested on two known benchmarks, FEI and Georgia Tech face datasets. Experimental results illustrate the robustness of the proposed algorithm.