There are many threats to vehicles that have autonomous driving capabilities. It turned out that the hacked boards, which created instant image projections, were also one of them.
As technology advances, driving assistant systems that help ensure drivers ‘ safety also continue to evolve. However, it is impossible to say that these systems are completely error-free.
In their study, security researchers found that roadside boards connected to the internet, which are interfered with from outside, can pose a serious security risk. Systems like Tesla’s Autopilot can therefore suddenly brake, change direction, or veer off the road.
Studies at Ben Gurion University of the Negev, located in Israel, have also focused on this type of problem. They successfully revealed that autonomous systems and driving assistance systems were open to experiencing problems because of the “ghostly” images.
By being able to instantly project non-existent images with captured boards, the researchers thus enabled automated systems to react to objects and images that actually did not exist at all.
In previous examples, images such as an image of a person were projected on roads and a stop sign on trees using the projection technique. Although the images appear only for a few milliseconds, for autonomous systems, this time is enough for them to detect the images. Advanced driver assistance systems think these images are real.
Although the same method was used in the new research, the images were projected from roadside boards, not from a projection. People who infiltrate these systems, which we have seen before, can cause cars to crash. And they can do it without leaving a trace.
The Mobileye 630 system, one of the latest Tesla Autopilot versions, was tested in this experiment. In the first version, 0.42 seconds was enough to fool the system, while in the second version, this time was reduced to 1 in 8 seconds.
If such attacks affect a larger number of vehicles, they can cause more accidents. The researchers ‘ findings will be presented at the ACM Conference on computer and Communications Security in November.