Ghost in a car shell: Engineers make self-driving vehicles ‘hallucinate’ at will – MadRadar is worrying proof-of-concept that should put car companies on alert
Engineers at Duke University, Durham, North Carolina, have developed a system that can manipulate radar sensors in cars and make vehicles “hallucinate” a series of scenarios. These include concealing the approach of an oncoming vehicle, making it appear as if a real car has suddenly changed direction, or creating a non-existent phantom car.
Called MadRadar, this system performs tasks quickly and requires no prior knowledge of the target vehicle’s specific radar settings.
“We can make a fake vehicle appear out of nowhere or a real vehicle disappear in real-world experiments,” said Miroslav Pajic, the Dickinson Family associate professor of electrical and computer engineering and leader of the team behind MadRadar. a statement.
Real world case studies
Radar is critical for modern vehicles equipped with assistive and autonomous driving technologies to detect nearby moving vehicles. Due to the variety of makes and models on the road, vehicles have slightly different operating parameters, which traditionally makes radar spoofing difficult.
However, MadRadar overcomes this hurdle by identifying a vehicle’s radar parameters within a quarter of a second and then initiating its own radar signals to deceive the targeted radar system.
Described as a “general black-box radar attack framework for automotive mmWave FMCW radars,” MadRadar can estimate the victim radar configuration in real time and then launch an attack based on the estimates.
The study states: “We evaluate the impact of such attacks that maliciously manipulate a victim radar point cloud, and demonstrate the novel ability to effectively ‘add’ (that is, false positive) attacks, ‘remove’ (that i.e. false negative attacks), or ‘move’ (i.e. translation attacks) object detections from the location of a victim vehicle. Finally, we experimentally demonstrate the feasibility of our attacks on real-world case studies performed using a real-time physical prototype on a software-defined radio platform.”
The researchers plan to detail their work at the 2024 Network and Distributed System Security Symposium in San Diego, California. The paper is currently available on the arXiv preprint server.
The researchers emphasize that the existence of MadRadar underlines the urgent need for manufacturers to improve the security measures of their radar systems to protect against potential misuse.
“We are not building these systems to hurt anyone, we are demonstrating the existing problems with current radar systems to show that we need to fundamentally change the way we design them,” Pajic concluded.