Attribution Using Keyboard Row Based Behavioural Biometrics for Handedness Recognition

In cases where an unknown intruder is using a computer we are often left attempting to attribute who they are. By using behavioural biometrics we can attempt to derive traits based on keystroke dynamics. Using hold times of keys, which we average and split into six groups based on rows and hand, we examine our ability to detect handedness. We present a novel method of identifying handedness and compare this to J48, Neural Network, and Random Forest machine learning algorithms. We find Random Forest the most effective method for handedness detection with an accuracy of 94.5% and a Kappa of 0.79. To our knowledge this is the best result achieved in handedness detection, and the only work using behavioural biometrics to identify handedness in long text.