Researchers Show They Can “Blind” Sensors in Self-Driving Cars

A team of Chinese and U.S. security researchers have been able to jam the sensor signals used by a self-driving car to detect its surroundings.

The researchers also  made self-driving systems “see” an object that isn’t there, make real objects vanish and temporarily blind a vehicle’s object-detecting cameras, Forbes reports.

The hacks were performed by an expert from Chinese security company Qihoo 360 and scientists at University of South Carolina and China’s Chejiang University. The researchers are presenting their results at this week’s Def Com hackers’ convention in Las Vegas.

The group was able to jam a Tesla Model S electric sedan’s Autopilot ultrasonic sensors, which are used for self-parking, with bogus signals that overpowered legitimate input. But the technique worked only at a distance of about three feet and required equipment costing more than $200,000.

The hackers also disrupted the Model S’s adaptive cruise control radar and “blinded” the vehicle’s object-detecting cameras. The tests were done with a stationary car, but the researchers describe the mere possibility of jamming sensors critical to a self-driving car’s performance as “worrisome.”

Forbes says the team also was able to jam ultrasonic sensors in Audi, Ford and Volkswagen vehicles. Tesla shrugged off the demonstration, noting that the techniques would require a very determined hacker with a very large budget.