Pattern representation using Neuroevolution of the augmenting topology (NEAT) on Keystroke dynamics features in Biometrics

Abstract

Since several decades, the standard authentication method of username and password is still termed as vulnerable. Biometrics technologies are therefore being implemented to provide the solution for human authentication. Among the various biometrics’ characteristics, Keystroke dynamics is widely used and easily accepted by users. It is the one which is smoothly deployed at the user end. Unlike other metrics adopted in other research works, the flight time, the distance between the keys and dwell time have been considered in this paper. With the rapid evolution of technology, many systems are becoming vulnerable due to different forms of attacks that can be interfered at different levels in the biometric system such as the attack on the matcher module and at the sensor level. Thus, to counteract these loopholes, the Neuroevolution of the Augmenting Topologies, which is a technique that can reduce the level of attack at the matcher module, have been applied to keystroke dynamics features and evaluated. Much effort has been placed in devising enhanced techniques to pre-process, represent and match features in the keystroke dynamics system. Firefly, yet another algorithm, which has achieved motivating results in feature subset selection has been enhanced and applied on the keystroke dynamics extracted features. Experiments carried on the proposed Neuroevolution of the Augmenting Topologies system have achieved a good recognition rate over other techniques like Chaotic Neural network and Neural network, which could produce only a recognition rate of about 86%.