Detecting Attacks Against Safety-Critical ADAS Based on In-Vehicle Network Message Patterns
Christopher N. Gutierrez, Marcio Juliato, Shabbir Ahmed, Manoj R. Sastry · 2019
Vehicles on the road today are Internet-enabled devices, providing navigation, safety, and entertainment to passengers. It is possible for an attacker to compromise these devices to gain remote access to the in-vehicle network, allowing control of the vehicle. To detect the presence of masqueraded messages, we propose a Message Time-series Intrusion Detection System (MTS IDS), which is based on the principle that ADAS messages exhibit regular (benign) patterns.We demonstrate the feasibility of detecting masquerade injection attacks that attempt to negatively influence Advanced Driver-assistance Systems such as Adaptive Cruise Control (ACC) and Lane Centering Systems (LCS). Our results show a MTS IDS detects masquerade messages against ACC and LCS ADAS systems with F1-scores of 0.9889 and 0.98705, respectively. While no computing system can be completely secure, these results can help increase resilience of ADAS and autonomous vehicles against masquerading attacks, and therefore improve road and passenger safety.