Prototype system wards off crashes, cyberattacks in autonomous vehicles

Video: College of Engineering

Researchers in the College of Engineering’s Department of Electrical and Computer Engineering have developed an intelligent transportation system prototype designed to avoid collisions and prevent hacking of autonomous vehicles.

Working in the Texas A&M University’s Cyberphysical Systems Laboratory, University Distinguished Professor P. R. Kumar and graduate students Bharadwaj Satchidanandan and Woo-Hyun Ko have applied the theory of dynamic watermarking of sensors in autonomous vehicles to prevent malicious attacks.

In their research demonstrations, 10 cameras recorded the movement of the self-driving prototype vehicles. The vision sensors in the system received the images and accurately calculated the exact location and orientation of the vehicles. Then they transmitted this information to a server, which in turn controlled the vehicles.

“Sensors are like GPS navigation in the network that gather information about the environment,” Satchidanandan said. “Actuators such as motors, or controls such as the steering wheel, interact with them. If the sensors are corrupted or hijacked by malicious agents through the Internet, they can provide false information on vehicle locations resulting in collisions.”

To fix this, Kumar and his team added a random private signal called a ‘watermark’ to the actuators. The watermark and its statistical properties were known to every node in the system, but its actual random values were not revealed. When the measurements reported by the sensors did not have the right properties of this watermark, the actuators assumed that the sensors or their measurements had been tampered with somewhere along the line. With this new information, the researchers could predict a collision.

The researchers showed that their technology could work in the lab. The actuators in the autonomous vehicles halted themselves when researchers tampered with the sensors.