Behavioural Biometric Authentication using Keystroke Features with Machine Learning

Abstract

Computers are depended on to store sensitive information and provide this information security from outsiders. Biometric authentication is very important in order to keep a system secure and keep information safe from outside attacks. Most of the biometric methods use expensive hardware and software in order to ensure this. Hence behavioural biometrics were proposed that perform authentication based on only behavioural characteristics of a user like gait, manner, way of typing. Keystroke dynamics offers great promise, emerging as one of the top ways for behavioural biometric authentication. Keystroke dynamics is extremely difficult to impersonate hence it is an extremely useful method for biometric authentication. In this paper, the best method for authentication using keystroke dynamics is discussed using the CMU Keystroke Dynamics Benchmark Data Set. Five Machine Learning Algorithms are compared to identify the best algorithm to be used to build the authentication system based on the lowest Equal Error Rate (EER).