Preventing battery attacks on electrical vehicles based on data-driven behavior modeling

Abstract

With the rapid development of wireless communication technologies for electrical vehicles (EVs), as a critical part of a pure EV, batteries could be attacked (e.g., draining energy) to reduce driving range and increase driving range anxiety. However, no methods have been proposed to ensure security of EV batteries. In this paper, we propose the first battery attacks, which can turn on air condition and stop battery charging process by sending requests through a smartphone without being noticed by users. We then propose a Battery authentication method (Bauth) to detect the battery attacks. We firstly build a data-driven behavior model to describe a user's habits in turning on air condition and stopping battery charging. In the behavior model, to distinguish users that share a vehicle for high modeling accuracy, we apply the random forest technology to identify each user based on battery state. Based on the established behavior model, we then build a reinforcement learning model that judges whether an AC-turn-on or batter-charge-stop request from a smartphone is from the real user based on current vehicle states. We conducted real-life daily driving experiments with different participants to evaluate the battery attack detection accuracy of Bauth. The experimental results show that Bauth can prevent EV batteries from being attacked effectively in comparison with another method and its attack detection accuracy reaches as high as 93.44%.