If someone hacks a self-driving car, the way it steers leaves a 'fingerprint' that's so weird the car can actually tell it's being hijacked.
arXiv · March 17, 2026 · 2603.14124
The Takeaway
Researchers found that different types of attacks—from digital network disruptions to 'phantom' images projected on the road—distort a car's steering and computer speed in distinct, predictable ways. These signatures are so specific that a car could identify exactly which kind of hack is occurring simply by monitoring its own mechanical jitters.
From the abstract
Deep learning-based perception pipelines in autonomous ground vehicles are vulnerable to both adversarial manipulation and network-layer disruption. We present a systematic, on-hardware experimental evaluation of five attack classes: FGSM, PGD, man-in-the-middle (MitM), denial-of-service (DoS), and phantom attacks on low-cost autonomous vehicle platforms (JetRacer and Yahboom). Using a standardized 13-second experimental protocol and comprehensive automated logging, we systematically characteriz