A Data-Driven Minimal Approach for CAN Bus Reverse Engineering

Abstract

Current in-vehicle communication systems lack security features, such as encryption and secure authentication. The approach most commonly used by car manufacturers is to achieve security through obscurity – keep the proprietary format used to encode the information secret. However, it is still possible to decode this information via reverse engineering. Existing reverse engineering methods typically require physical access to the vehicle and are time consuming. In this paper, we present a Machine Learning-based method that performs automated Controller Area Network (CAN) bus reverse engineering while requiring minimal time, hardware equipment, and potentially no physical access to the vehicle. Our results demonstrate high accuracy in identifying critical vehicle functions just from analysing raw traces of CAN data.