Researchers say they can trick self-driving cars into misidentifying road signs

eu stars
IRL UK



Researchers say they can trick self-driving cars into misidentifying road signs

Researchers from the University of Washington have identified a very simple way to trick self-driving cars into misidentifying road signs.

They have found that by strategically placing stickers on different road signs, they can effectively fool the image-processing software in autonomous cars.

In one dangerous experimental incident, stickers attached to a stop sign caused the onboard sensors to misidentify it as a speed-limit sign.

If this was a real world situation, the car may not have observe the stop sign with cataclysmic results.

Security researchers basically performed an analog hack, which means they did not need to access and manipulate the onboard digital computer. Instead of sending illicit commands digitally, the researchers simply examined the program used by the car to identify different objects, they were able to create printable stickers based on known faults in how the system processes images.

Most autonomous-driving systems use onboard cameras to compare what the car is "seeing" with stored images in the computer, so changing the appearance of an object can cause the system to make a mistake.

Researchers have been aware of this particular problem with autonomous systems for some time. In India, for example, engineers have had trouble getting self-driving cars to identify the three-wheeled auto-rickshaws, as drivers have altered the vehicles in so many ways that they become unidentifiable to the autonomous car sensors.

The University of Washington researchers have said that this glitch in the object identifying system could be used to cause havoc in the real world.

The stickers are effective enough to work in most conditions, but subtle enough that they may not immediately be noticed by human drivers or the police. The stickers could even mess with the speed-limit recognition systems already available on some production cars.

Ultimately, self-driving cars may need to develop the very human ability of identifying variations in an object's appearance. 

A startup called Cortica is developing so-called "unsupervised machine learning" for self-driving cars, which could give them that ability, something we all take for granted.

Dryve.com, for example, provide updated news regarding driver-less cars and their future. You may like to check out their infographic for the future of driverless cars.

(As always, if you or a family member are considering buying a used car, don’t buy until you run a car check report with MyVehicle.ie where you will find out the true history of the vehicle.)









Author

Justin Kavanagh
Justin Kavanagh is a recognised leader in automotive intelligence and vehicle data supply to the entire motor industry. He has almost 20 years experience in building systems from the ground up. As the Managing Director of Vehicle Management System, he understands the need and importance of trustworthy and reliable vehicle history and advice to both the trade and the public.
Follow me on LinkedIn

eu stars
IRL UK