Context Aware Anomaly Detection in Automotive CAN Without Decoding

Explore a cutting-edge conference talk that addresses critical cybersecurity vulnerabilities in modern automotive systems through innovative anomaly detection techniques. Learn how traditional security measures fail against sophisticated attackers who can mimic normal sensor behaviors while manipulating vehicle states, and discover a revolutionary approach that leverages inter-signal correlations without requiring CAN bus decoding or reverse engineering. Master deep sequence-learning methodologies specifically designed for raw CAN payloads, focusing on time-aware and context-sensitive detection patterns that can identify stealthy attacks in real-time cyber-physical automotive environments. Understand how exploiting timing patterns and contextual relationships between signals provides superior security compared to conventional identifier-based detection systems. Witness practical applications through live demonstrations using authentic CAN datasets and emulated automotive environments, showcasing how this approach redefines trust and security in connected vehicles where even minor manipulations can result in catastrophic consequences.