October 16, 2023 at 10:36AM
Australian researchers have developed an AI-driven cyber intrusion detection system to assist unmanned military robots in identifying man-in-the-middle (MitM) cyberattacks. The system, which uses deep learning convolutional neural networks (CNNs), aims to reduce vulnerabilities in the robot operating system (ROS) used by civilian and military robots. The algorithm achieved 99% accuracy when tested on a US Army ground vehicle. The researchers highlight that ROS vulnerabilities make the robots susceptible to various cyberattacks, including breaches, hijacking, and denial-of-service (DoS).
The meeting notes discuss a new cyber intrusion detection system developed by two Australian academic researchers. The system relies on AI and deep learning convolutional neural networks (CNNs) to help unmanned military robots detect man-in-the-middle (MitM) cyberattacks. The researchers conducted tests on a US Army ground vehicle, and the algorithm achieved a 99% accuracy rate. The cyber-intrusion detection framework primarily focuses on MitM attacks but also addresses vulnerabilities in the robot operating system (ROS) that make it susceptible to breaches, hijacking, denial-of-service (DoS), and other types of cyberattacks. The researchers highlight that robotic systems have multiple levels of compromise and preventing these attacks is challenging, especially for complex and modern robots. The MitM-detection algorithm was tested on ground robots connected to separate computers over a Wi-Fi network, and the cyberattack caused the robot to become unresponsive by overriding the guidance signal with unintended traffic data. The algorithm was trained using collected data under legitimate and cyberattack conditions, resulting in high accuracy and performance compared to other detection techniques. The researchers also plan to test their algorithm on different robotic platforms, including unmanned aerial vehicles.