VSSIDS: A Novel Intrusion Detection Scheme Based on Signal Stretching for In-Vehicle Networks

Over the past decade, voltage-based intrusion detection systems (IDSs) have faced a persistent and unresolved challenge: models trained by voltage signals collected within a narrow temperature range often fail to generalize effectively to environments with wider temperature variations. This limitation stems from two primary reasons. The voltage characteristics of controller area network bus signals exhibit noticeable variations under different environmental temperatures. Furthermore, although it is feasible to obtain the current environmental temperature of the vehicle, the actual operating temperature of each electronic control unit is difficult to determine. In this paper, we propose a novel scheme named VSSIDS, which responds to the challenge by stretching all training and testing signals to the same amplitude. This mitigates the impact of temperature variations on voltage signals, allowing the trained model to remain effective across environments with wider temperature variations. Our experiments on both a real vehicle and a prototype demonstrate the effectiveness of VSSIDS in intrusion detection and its robustness against temperature variations.