Field Testing and Detection of Camera Interference for Autonomous Driving

TL;DR

This paper addresses camera interference attacks in automotive ethernet-driven IVNs and presents a GRU-based intrusion detection system that operates on time-series features derived from packet length sequences using a sliding-window approach. The method is validated on a commercial vehicle (Hyundai Genesis G80) with H.264 FU-A fragmentation, achieving an AUC of 0.9982 and a true positive rate of 0.99, demonstrating strong discrimination between normal and attacked traffic. It contributes a real-vehicle CIA demonstration, a dedicated intrusion dataset with normal and attack traffic, and a GRU-based IDS tailored for automotive ethernet, while acknowledging limitations related to fragmentation method and encoding, and proposing payload-recovery-based detection as future work.

Abstract

In recent advancements in connected and autonomous vehicles (CAVs), automotive ethernet has emerged as a critical technology for in-vehicle networks (IVNs), superseding traditional protocols like the CAN due to its superior bandwidth and data transmission capabilities. This study explores the detection of camera interference attacks (CIA) within an automotive ethernet-driven environment using a novel GRU-based IDS. Leveraging a sliding-window data preprocessing technique, our IDS effectively analyzes packet length sequences to differentiate between normal and anomalous data transmissions. Experimental evaluations conducted on a commercial car equipped with H.264 encoding and fragmentation unit-A (FU-A) demonstrated high detection accuracy, achieving an AUC of 0.9982 and a true positive rate of 0.99 with a window size of 255.

Figures (8)

• Figure 1: Overview of the proposed methodology for intrusion detection workflow.
• Figure 2: Example for attack autonomous driving module. (A) is ethernet switch, (B) is infotainment display, (C) is autonomous driving module, D.1, D.2, D.3, D.4 are on-board camera.
• Figure 3: Cache poisoning attack in vehicle.
• Figure 4: Demonstration of the camera inference attack. The infotainment system becomes blind when it goes under the attack.
• Figure 5: Heatmap for length sequence 2-gram feature.