Blinding Autonomous Vehicles with Lasers

Reading Time: 3 minutes
Source: Sara Rampazzi/University of Florida

Autonomous vehicles, a medium of transport that many believe will become the standard in the future. In recent years, the progress in the development of this type of technology has risen exponentially. Self-driving cars are already superior to humans in certain abilities that affect the safety of transportation. Primarily their faster reaction time. They also do not suffer from hindered focus due to tiredness, distractions or use of alcohol like we do.

However, autonomous vehicles are still inferior to human drivers when it comes to recognizing the situation on the road. There are three main sensor systems through which self-driving cars collect the information regarding the space around them: camera, radar, and LiDAR systems.

LiDAR stands for Light Detection and Ranging. It works by sending a laser light, capturing the reflected light from objects around it and then, calculating the distance to said objects thanks to the recorded time of flight of the laser light. Its widely believed to be crucial in making cars fully autonomous in the future.

Thanks to researchers from the University of Florida, the University of Michigan and the University of Electro-Communications in Japan this sensor can be improved through their discovery of an alarming error that can be maliciously caused by third parties.

Source: Sara Rampazzi/University of Florida

They were first to discover that a laser light aimed at a LiDAR system can imitate the systems own reflected LiDAR laser creating a blind spot in its virtual map of objects. The created blind spot can erase obstacles, crossing pedestrians or even other vehicles from its view. It is not hard to imagine how that could lead to tragic consequences on the road.

In their experiment they setup the third-party laser 4.5 meters from the theoretical road aimed at a stationary LiDAR sensor on top of a vehicle. A person walked in front of the vehicle providing data for the system. After analyzing the LiDARs map of objects data, they were able to find that the sensor had no problem with detecting the pedestrian at the start. But, as soon as he walked into the 8 ° range of the attack region he was immediately removed from the systems view. Upon leaving the range of the blind spot the sensor was again able to detect the pedestrian in its view.

The attack can be recreated by others without much difficulty with the tracking of the LiDAR system on top the car being the hardest part. The choice of a laser that can emulate the sensors reflected lights is not much trouble as LiDAR sensor producers publish their sensors technicalities publicly.

As troubling as this discovery may seem. I am personally grateful that this type of vulnerability has been exposed that early into the development of self-driving cars. Giving us plenty of time for improvement before they freely roam our streets. A future where movie villains could just aim a laser at a car approaching their target is not something that any of us would want to be a part of.


Bibliography:

Yuen, Desmond. “Can You React Faster than a Self-Driving Car on 5G Networks?” MEDIUM. Published January 30, 2021. https://medium.com/predict/making-roads-safer-with-self-driving-cars-and-5g-c1e28526362c

University of Florida. “Laser attack blinds autonomous vehicles, deleting pedestrians and confusing cars” TechXplore. Published October 31, 2022. https://techxplore.com/news/2022-10-laser-autonomous-vehicles-deleting-pedestrians.html

Cao, Yulong. Bhupathiraju, S. Hrushikesh. Naghavi, Pirouz. Sugawara, Takeshi. Mao, Z. Morley. Rampazzi, Sara. „You Can’t See Me: Physical Removal Attacks on LiDAR-based Autonomous Vehicles Driving Frameworks.” arXiv. Last revised October 27, 2022. https://arxiv.org/abs/2210.09482

Synopsys. “LiDAR.” Synopsys. Accessed December 9, 2022. https://www.synopsys.com/glossary/what-is-lidar.html

Leave a Reply

WordPress › Error

There has been a critical error on this website.

Learn more about debugging in WordPress.