The emergence of self-driving vehicle networks that collaborate and communicate with each other or with infrastructure to make decisions has brought forth a new set of challenges. According to a recent study led by the University of Michigan, these networks are vulnerable to data fabrication attacks. This study, presented at the 33rd USENIX Security Symposium in Philadelphia, sheds light on the risks associated with collaborative perception in connected and autonomous vehicles.
The concept of collaborative perception in vehicle-to-everything (V2X) networks allows connected and autonomous vehicles to enhance their capabilities by leveraging shared information. However, this collaboration opens up opportunities for malicious actors to manipulate data. Hackers could introduce fake objects or alter real objects in the perception data, potentially causing vehicles to make dangerous decisions such as hard braking or even collisions.
The study conducted by researchers at the University of Michigan focused on understanding and countering these attacks. Unlike previous studies that examined individual sensor security, this research introduced sophisticated, real-time attacks in both virtual simulations and real-world scenarios at U-M’s Mcity Test Facility. By administering falsified LiDAR-based 3D sensor data with malicious modifications, the researchers were able to test the effectiveness of these attacks on connected and autonomous vehicles.
To address these security vulnerabilities, the researchers developed a countermeasure system called Collaborative Anomaly Detection. This system leverages shared occupancy maps, which are 2D representations of the environment, to cross-check data and detect geometric inconsistencies in abnormal data. In virtual simulated environments, the system achieved a detection rate of 91.5% with a false positive rate of 3%, demonstrating its effectiveness in reducing safety hazards.
The findings of this study provide a solid framework for improving safety in connected and autonomous vehicles. By detecting and countering data fabrication attacks in collaborative perception systems, this research not only advances vehicle security but also safeguards passengers and other drivers. Moreover, by offering benchmark datasets and open-sourcing their methodology, the researchers hope to set a new standard for future research in this field, encouraging further innovation in autonomous vehicle safety and security.