Protecting connected, self-driving vehicles from hackers

Emerging networks of self-driving vehicles that collaborate and communicate with each other or with infrastructure to make decisions are vulnerable to data falsification attacks, according to a University of Michigan study that also outlines preventative measures for fleet operators.

The researchers recently presented the work at the 33rd USENIX Security Symposium in Philadelphia. The paper is published on the arXiv Preprint server.

Although this network of collaboration and communication, known as Vehicle-to-Everything, or V2X, is not yet on the roads, many countries are supporting the development of this technology and have begun small-scale testing. The U.S. Department of Transportation recently released a V2X deployment plan to guide implementation of the technology as it moves forward.

“Through collaborative sensing, connected and autonomous vehicles can ‘see’ more than they could alone by leveraging the collective sensor power and data insights of a network of vehicles. But this power comes with serious safety risks,” said Z. Morley Mao, a professor of computer science and engineering at UM and lead author of the study.

By exchanging information between vehicles, hackers can introduce fake objects or remove real objects from the perception data. This can lead to heavy braking or accidents of the vehicles.

“Understanding and defending against attacks is a critical step forward not only to improve the security of connected and autonomous vehicles, but also to protect passengers and other drivers,” said Qingzhao Zhang, a doctoral student in computer science and engineering at UM and lead author of the study.

While previous studies focused on the security of individual sensors or simpler collaboration models, this study presented sophisticated real-time attacks tested in both rigorous virtual simulations and real-world scenarios at UM’s Mcity Test Facility, a test site for connected and automated vehicles and technologies.

To understand security vulnerabilities, the researchers ran fake LiDAR-based 3D sensor data that appears realistic to the system but contains malicious modifications through physical access to the hardware and software system. They used zero-delay attack planning, a high-risk cyberattack that uses precise timing to introduce malicious data without any delay or lag.

In virtual simulation scenarios, the attacks were extremely effective with a success rate of 86%. Attacks on the road against three vehicles in the Mcity area triggered collisions and emergency braking.

The countermeasure system, called Collaborative Anomaly Detection, uses shared occupancy maps – two-dimensional representations of the environment – ​​to verify data, allowing vehicles to quickly identify geometric inconsistencies in abnormal data.

The system achieved a detection rate of 91.5% with a false positive rate of 3% in virtually simulated environments and reduced security risks in the Mcity scenarios.

The findings provide a robust framework not only for improving the security of connected and autonomous vehicles, but also for detecting and defending against data falsification attacks in collaborative perception systems used in transportation, logistics, smart city initiatives or defense.

“By providing comprehensive benchmark datasets and making our methodology open source, our study sets a new standard for research in this area and promotes further development and innovation in the field of autonomous vehicle safety,” Mao said.