Self-driving cars deceived by billboards: here's how

Source: HW Upgrade added 19th Oct 2020

Digital billboards could be exploited as an attack vehicle to cause accidents and traffic jams, forcing self-driving cars to perform sudden maneuvers

by Andrea Bai published on , at 08: 42 in the Technology channel

The development activities within the autonomous driving have carried out and continue to make significant progress, but it is undeniable that there is still a long way to go (in every sense!) before this technology can be considered completely autonomous and with a degree of safety that can be used without requiring the supervision of a human driver.

The field experimentation of this technology has resulted in some unpleasant episodes (the best known is the fatal accident that involved in a woman hit by an Uber self-driving car) and in this phase the focus is on the weather conditions, the mapping systems and how the cars should behave in relation to dangerous objects or situations on the road.

Ghost images on a billboard, and the car suddenly brakes

A group of security researchers from Ben Gurion University of the Negev, however, conducted a work that was able to demonstrate how self-driving cars could be deceived and induced to carry out certain actions through “compromised” billboards. Researchers call them “ghost images” that humans may not detect with the naked eye and give the example of a stop sign that can be created with flickering lights on an electronic billboard, thus causing the car to suddenly stop. .

Yes it would therefore be light projections lasting a few frames, but sufficient to make the car “brake or steer”, without the driver being able to understand what was the cause of the sudden behavior of the car. The researchers performed some tests with cars equipped with Tesla’s Autopilot and with MobileEye . According to the researchers, a phantom stop sign lasting 0, 42 has deceived the ‘Autopilot by Telsa, while a signal lasting 1/8 of a second was sufficient for MobileEye.

The experiments conducted by researchers at Ben Gurion University are based on other previous research that used light projections of a few fractions of a second creating silhouettes of human beings, in order to confuse the autonomous vehicles on the road. These are tests from which similar results have emerged, but the ability to exploit this behavior with digital billboards is a more dangerous prospect since hackers and criminals could easily have a way to cause accidents and traffic jams, with the risk of bringing an entire city to its knees without a trace .