Researchers at the University of South Australia and Charles Sturt University have developed an algorithm designed to detect and prevent man-in-the-middle (MitM) attacks on unmanned military robots. MitM attacks involve intercepting data traffic between two parties, such as a robot and its legitimate controllers, for the purpose of eavesdropping or injecting false data into the stream.
These attacks can disrupt the operation of unmanned vehicles, modify instructions, and even assume control of the robot, leading to potentially dangerous actions. The researchers used machine learning techniques to develop an algorithm that successfully detected and thwarted these attempts in a replica of the GVR-BOT used by the U.S. Army (TARDEC), with a 99% success rate in preventing attacks and minimal false positives.
Unmanned military robots are increasingly networked and susceptible to cyberattacks due to their reliance on the robot operating system (ROS) and the need for collaborative work within the context of Industry 4.0. The algorithm developed by the researchers analyzes network traffic data, using node-based methods, packet data scrutiny, and flow-statistic-based systems to detect any attempts to compromise the robot.
The algorithm employs a deep learning convolutional neural network (CNN) model, making it effective in identifying cyberattacks even after just a few epochs of model training. Future applications for this intrusion detection system could extend to more complex robotic platforms, such as unmanned aerial vehicles, with optimized versions of the protection system.
