Australian researchers have used AI to detect and shut down man-in-the-middle (MITM) attacks on the popular open source Robot Operating System (ROS).
GVR-BOT
University of South Australia
As they explain in this IEEE publication, the researchers conducted a pentest on the American military robot, GVR-BOT, which runs ROS on its onboard computer, to discover the vulnerabilities.
They collected network data of the robot operating under “legitimate and malicious conditions” and used that to train a convolutional neural network (CNN) to try and identify attack traffic.
UniSA researcher Anthony Finn and Dr Fendy Santoso from the Charles Sturt University AI and Cyber Futures Institute worked with the US Army Futures Command to replicate the MITM attack.
However, they say the algorithm their CNN produced, running on ROS, “was 99 percent successful in preventing a malicious attack,” according to this UniSA statement.
“False positive rates of less than two percent validated the system,” the university said.
“The robot operating system (ROS) is extremely susceptible to data breaches and electronic hijacking because it is so highly networked,” Professor Finn said.
In a world where sensors, actuators and controllers need to exchange information over the cloud, such systems are “highly vulnerable to cyberattacks”, he said.
ROS, in particular, is vulnerable because it relies on encrypted network traffic as its main defence.
The operating system “largely ignores security issues in its coding scheme”, the university said, and has “limited integrity-checking”.
The pair plan to test their intrusion detection algorithm on different robotic platforms, such as drones, whose dynamics are faster and more complex compared to a ground robot.
The research was published in IEEE Transactions on Dependable and Secure Computing.