Hah! I was just talking about self-driving cars with the other (read: better) half! Ok, I may have said something like, “Didn’t ‘Demolition Man’ take place in 2023 or something? I mean, we should have incredibly capable self-driving cars by now.” His response: “Well, if that’s the case, we should be well past that. Back to the Future? Hello!” I still haven’t found out the three sea shell thing and the Verbal Morality Statute earning f-word is still my favorite, so I guess it’s a good thing I was dyslexic about the year. It was 2032. Whew! I’m safe!
Turns out self-driving cars are far from “incredibly capable” at the moment. These cars use lidar – a laser-based radar kind of system. The lidar creates a 3-D environment for the self-driving car to make its driving decisions. Tech lesson one over.
Jonathan Petit, Principal Scientist at Security Innovation, spoke of his research to IEEE Spectrum. Using a $60 setup comprised of a low-power laser and a pulse generator, he was able to, essentially, disable an IBEO Lux lidar system. He had to synchronize his gizmo, which he said was “the tricky part.” Petit acknowledged his attacks are currently limited to the one specific unit, but suggests none of the lidar manufacturers have thought about or tried this type of experiment.
Petit was able to create illusions of fake cars, walls, and pedestrians, and even make them move. Once the car’s lidar detects something like that, the car will slow down or even come to a complete stop to avoid hitting the obstacle. Tech lesson two complete.
The experiment worked as far as 100 meters away (approx. 348 feet) from all directions, and didn’t necessarily require precise targeting from a narrow beam. It’s also important to note the lidar systems are not encrypted, but this issue can be solved before it actually becomes a problem.
I sure hope they get this patched before it does some real damage. What about you? How do you feel about a) self-driving cars and/or b) knowing they can be confused like this?