EBAT: Evaluation of Behavioural Authentication Toolbox

Sensors, ubiquitously present in our surroundings and everyday objects, can reflect our behaviour regarding these spaces and objects and, consequently, open opportunities for novel behaviour-based authentication methods. Despite the potential for a non-intrusive, privacy-aware, and continuous operation that behaviour-based authentication holds, it is yet to make the leap “out of the lab”. A key obstacle to this is the lack of a framework that would enable objective comparison between different authentication approaches proposed within the research community. In this paper, we introduce EBAT, an Evaluation of Behavioural Authentication Toolbox, which encompasses the most commonly used, some recently proposed, and some newly developed evaluation metrics, as well as an evaluation framework specifically tailored for nuances of sensor-based behaviour authentication. We implement EBAT as an easy-to-integrate software package and demonstrate its usability by assessing a number of state-of-the-art behavioural authentication methods.