In-vehicle device data tampering detection: Accurate identification based on correlation calculation and data relationship

Abstract

The rapid advancement of intelligent connected vehicles (ICVs), driven by the integration of AI and 5G, has intensified the need for reliable accident forensics. We present a novel correlation analysis-based method for detecting tampered vehicle electronic data, addressing critical security vulnerabilities in current systems. Our approach establishes multivariate relationship clusters from in-vehicle data characteristics, performs dimensionality reduction, and computes anomaly scores through tail probability analysis. The experimental results demonstrate that the proposed method exhibits superior detection performance compared to existing approaches for random injection attacks, targeted tampering attacks, and outlier attacks.