Machine Learning Algorithm on Keystroke dynamics Fused pattern in biometrics
Purvashi Baynath, K. M. S. Soyjaudah, Maleika Heenaye-Mamode Khan · 2019
Nowadays, Multimodal biometric is being deployed to overcome the challenges of spoofing which is encountered by unimodal biometrics. However, security is still compromised using multi modal biometric, thus requiring the attention of researchers. In view to increase security, an attempt of deploying multiple machine learning techniques on fused pattern have been explored and developed. Two enhanced algorithms of Chaotic Neural Network and NeuroEvolution of the augmenting topology have been devised and applied on the fused features. To verify the trustworthiness of the proposed system, several experiments have been deployed on both unimodal and multimodal systems. After several experiments performed, the algorithms reveal that multi biometric system exhibit higher recognition rates and lower false rejection rates compared to unimodal systems. In this research, we have also shown that algorithms have the possibility to countermeasure the loopholes that may occur at the matching module. We have achieved a good recognition rate of 99.1% by adopting score level fusion by using the proposed NeuroEvolution of the augmenting topology algorithm compared to previous research work conducted. Thus, the proposed solution has proved to be stable while using fused features.