Performance Evaluation of Fingerprint Dynamics in Machine Learning and Score Level Fusion Framework

Abstract

ABSTRACTRecently, rising trend of biometric-based authentication, like fingerprint recognition, in portable devices drawn the attention of researchers to examine the impending threats and inadequacies. With copious spoof attacks being reported for the traditional fingerprint and other biometric systems, assistive biometric technologies have emerged as a convenient way for enhancing the robustness of the user authentication systems. This paper examines a novel assistive biometric technique to fingerprint recognition, namely fingerprint dynamics, which involves the recording of time events during fingerprint acquisition. We investigate the potential of fingerprint dynamics as a standalone identification tool by comparing it with well-established analogues technique, keystroke dynamics. A sample data-set with several hundred biometric scans has been prepared from 19 users, over a period of six weeks to incorporate the dynamics comprehensively. Experimental results affirm the competency of the proposed technique over the work in comparison. Further, we have proposed score level fusion of support vector machine and artificial neural network classifiers in order to jointly account their performances. The empirical results exhibited up to 5.3% and 8.3% improvement in identification accuracies for fingerprint dynamics and keystroke dynamics respectively with the proposed fusion scheme.