Autonomous vehicles Cybersecurity

McAfee Investigation finds autonomous cars can be hacked through AI system

A new investigation conducted by McAfee has found that connected cars may be at risk of being hacked through the vehicles’ artificial intelligence (AI) system.

The method, which the company has called “model hacking” allows a person to over-ride and control a car’s software resulting in major safety concerns, along with creating an unclear future for autonomous vehicles. This process can currently compromise any ADAS software that is currently on the market including that installed in Tesla cars.

A team from McAfee proved this theory by conducting an attack on the black-box of the Mobileye EyeQ3 camera system on a Tesla Model S. The hack demonstrated how manipulating the AI system can cause the system to misinterpret information, in this case, a sign indicating a speed limit of 35 mph and causing the vehicle to travel at 85 mph.

The manipulation works by changing the AI’s algorithm that defines the parameters of the system and altering them in a negative way. By changing the training set that created the AI platform, hackers are able to seize control of nearly every aspect of the system.

While there are yet to be any reports of model hacking in real life, as the number of autonomous vehicles rapidly increases, so too does the risk. A Gartner report forecast “more than 740,000 autonomous-ready vehicles to be added to the global market in 2023.”

Researches on the study are optimistic about the results saying, “The good news is that much like classic software vulnerabilities, model hacking is possible to defend against, and the industry is taking advantage of this rare opportunity to address the threat before it becomes of real value to the adversary.”