Hackers have been able to trick many Tesla cars into increasing their speed by 50 kilometers per hour. Specifically, researchers tried to deceive a car’s MobilEye EyeQ3 camera system by changing the signal to reflect the maximum speed limit at the end of a road in such a way that a person driving near it would almost never notice it.
The result of this research conducted by McAfee’s cyber security firm is further proof that machine learning can destroy autonomous driving systems, thus posing a safety challenge for those who seek to commercialize technology.
Two members of McAfee’s Threat Research Team, Steve Povolny and Shivangee Trivedi , said that MobilEye EyeQ3 camera systems “read” speed limits signals and feed information into autonomous driving features, such as auto-driving features. Tesla speed control system.
The researchers stuck to a sign indicating the speed limit of such a tiny sticker that goes unnoticed. The camera read that the signal showed a maximum speed limit of 85 instead of 35, resulting in Tesla’s Model X 2016 and Model S of the same year increasing their speed by 50 kilometers per hour.
This demonstrates the fact that machine learning systems can be “hacking” and being “fooled” into critical and dangerous situations. In 2019, hackers managed to fool a Tesla car into a wrong lane, putting stickers on the road to manipulate the machine’s learning algorithms. As autonomous systems multiply, the issue extends to machine learning algorithms far beyond vehicles. A March 2019 study showed that medical machine learning systems were “fooled” and misdiagnosed.
McAfee’s research was made known to Tesla and MobilEye EyeQ3. Tesla, however, declined to comment on MIT Technology Review’s request. He said, however, that he had acknowledged McAfee ‘s findings and that he did not intend to correct issues with that generation of hardware . A Mobileye spokesman undermined the investigation by saying that the modified mark could even “fool” a man into reading 85 instead of 35. The company does not consider cheating on the camera an attack, and said it despite the role. that the camera plays on Tesla’s “cruise control” was not designed for autonomous driving.
A spokesman for Mobileye said autonomous vehicle technology would not only be based on sensation, but would be supported by various other technologies and data, such as crowd-mapping, to ensure the reliability of information received by camera sensors and to offer more security.
Since then, Tesla has decided to incorporate proprietary cameras into the newer models of its cars, while MobilEye EyeQ3 has released several new versions of its cameras, which in initial tests did not appear to be prone to attack. Finally, Povolny said there is still a large number of Tesla cars running on vulnerable hardware, while also pointing out that Tesla cars that have the first hardware version cannot be upgraded to newer hardware.