Hackers have demonstrated that it is possible to deceive the sensors in the Tesla Model S Autopilot system, with potentially disastrous consequences.
Tesla’s Autopilot uses three methods to detect its surroundings: radar, ultrasonic sensors and cameras. Researchers Chen Yan, Jianhao Liu and Wenyuan Xu found ways to attack all of these.
In their proof-of-concept demonstration, the researchers demonstrated how they were able to jam the car’s radio signals by placing two pieces of radio equipment (a signal generator and a VDI frequency multiplier) on a cart in front of the car, in order to simulate another vehicle. This “car” was indeed picked up by the system, however once the team switched on radio interference, it became invisible. If performed in a real life situation, this hack could cause the Tesla to crash into the invisible vehicle or object.
It is worth noting however, that the equipment used in this hack was highly expensive and according to the researchers, the radar attack would have to be aimed at precisely the right angle to hit a moving Tesla’s radar sensor. Therefore, it could prove difficult to achieve in real life.
Alternately, the researchers were able to convince the system to avoid objects that were not actually there. This involved targeting the car’s short range ultrasonic sensors, which are used for self-parking. To trick the sound-based sensors, they used a function generator for creating particular voltages and an ultrasonic transducer to convert the electricity to sound waves. This noise confused the sensors and was able to convince the car that it was not safe to park in a certain spot because there was an obstacle in its way. This method was also able to jam the sensors, causing them to ignore actual objects.
The researchers also found that they were able to render objects invisible to the car by simply wrapping them up in black foam.
“[Tesla] need[s] to think about adding detection mechanisms,” said Xu in a comment to Wired.
“If the noise is extremely high, or there’s something abnormal, the radar should warn the central data processing system and say ‘I’m not sure I’m working properly’.
“…Highly motivated people could use this to cause personal damage or property damage. We hope people get from this work that we still need to improve the reliability of these sensors … And we can’t simply depend on Tesla and not watch out for ourselves.”
Tesla responded positively to the research, according to Yan.
“They appreciated our work and are looking into this issue,” he said to Business Insider Australia.