Algorithm protects robots from malicious cyberattacks in seconds
19 October 2023
Image: University of South Australia
Australian researchers have devised a new algorithm capable of swiftly neutralising man-in-the-middle (MitM) cyberattacks targeted at unmanned military robots, achieving an impressive 99 percent success rate in real-time tests.
In an experiment using deep learning neural networks to simulate the behaviour of the human brain, artificial intelligence experts from Charles Sturt University and the University of South Australia (UniSA) trained the robot's operating system to learn the signature of a MitM eavesdropping cyberattack. This is where attackers interrupt an existing conversation or data transfer.
The algorithm, tested in real-time on a replica of a United States Army combat ground vehicle, was 99 percent successful in preventing a malicious attack. False positive rates of less than two percent validated the system, demonstrating its effectiveness.
UniSA autonomous systems researcher, Professor Anthony Finn, says the proposed algorithm performs better than other recognition techniques used around the world to detect cyberattacks.
Professor Finn and Dr Fendy Santoso from Charles Sturt Artificial Intelligence and Cyber Futures Institute collaborated with the US Army Futures Command to replicate a man-in-the-middle cyberattack on a GVT-BOT ground vehicle and trained its operating system to recognise an attack.
"The robot operating system (ROS) is extremely susceptible to data breaches and electronic hijacking because it is so highly networked," Professor Finn says.
"The advent of Industry 4.0, marked by the evolution in robotics, automation, and the Internet of Things, has demanded that robots work collaboratively, where sensors, actuators and controllers need to communicate and exchange information with one another via cloud services.
"The downside of this is that it makes them highly vulnerable to cyberattacks.
"The good news, however, is that the speed of computing doubles every couple of years, and it is now possible to develop and implement sophisticated AI algorithms to guard systems against digital attacks."
Dr Santoso says despite its tremendous benefits and widespread usage, the robot operating system largely ignores security issues in its coding scheme due to encrypted network traffic data and limited integrity-checking capability.
"Owing to the benefits of deep learning, our intrusion detection framework is robust and highly accurate," Dr Santoso says. "The system can handle large datasets suitable to safeguard large-scale and real-time data-driven systems, such as ROS."
Professor Finn and Dr Santoso 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.