A group of researchers developed an attack to trick “Level 0” autonomous vehicle sensors by using drones which project images too quick for humans to see but slow enough for the vehicle’s sensors.
Level 0 autonomy systems advise human drivers but don’t directly operate the vehicle and Ben Gurion University security researchers performed the experiment on a Renault Captur's system and were able to spoof various traffic signs including one that tricked the car’s sensors into believing false speed limits, according to the report.
Although the experiments posed no immediate threat with the Level 0 autonomy system, researchers stressed that the attacks would have greater consequences had they been carried out against more advanced systems and could make vehicles swerve, travel at unsafe speeds, and detour into unsafe territory.
To defend against these sort of attacks, researchers recommend cross-correlating input form multiple sensor suites to detect anomalies is to use good defense-in-depth and ensure the defense layers are robust enough to accurately detect anomalies.
