Correlation-Based Anomaly Detection for the CAN Bus

Abstract Previous attacks have shown that in-vehicle networks vulnerabilities and a successful attack could lead to significant financial loss danger life. In this paper, we propose Pearson correlation based anomaly detection algorithm detect CAN message modification attacks. The does not need priori information about the communication: it identifies signals on statistical properties, finds important coefficients for correlating signals, detects as deviations from previously learned normal state.