Professors from the University of South Australia and Charles Sturt University have collaborated to create an algorithm designed to identify and prevent man-in-the-middle (MitM) attacks targeting unmanned military robots.
MitM attacks constitute a form of cyberattack wherein data traffic between two parties, in this instance, the robot and its authorized controllers, is intercepted with the intent to eavesdrop or insert false data into the communication stream.
These malicious attacks aim to disrupt the functioning of unmanned vehicles, alter transmitted instructions, and sometimes even assume control, directing the robots to undertake hazardous actions.
Professor Anthony Finn, who was involved in the study, highlighted, "The robot operating system (ROS) is extremely susceptible to data breaches and electronic hijacking because it is so highly networked."
"The rise of Industry 4, characterized by advancements in robotics, automation, and the Internet of Things, has necessitated that robots operate collaboratively, where sensors, actuators, and controllers need to communicate and exchange information with one another via cloud services. However, this heightened connectivity makes them highly susceptible to cyberattacks."
The research team from the universities developed an algorithm incorporating machine learning techniques to recognize such attempts and thwart them within seconds.
The algorithm underwent testing using a replica of the GVR-BOT employed by the U.S. Army (TARDEC) and effectively prevented attacks in 99% of the instances, with false positives occurring in less than 2% of the trials.
Detecting MitM:
Detecting MitM attacks on unmanned vehicles and robots is a complex task due to their operation under fault-tolerant modes, making it challenging to differentiate between regular operations and fault conditions.
Moreover, robotic systems can be compromised at various levels, from the core system to sub-systems and their sub-components, potentially causing operational issues and rendering the robot non-functional.
To tackle this problem, the research team established a system that analyzed the network traffic data of the robot to identify any compromise attempts. The system employed node-based techniques, analyzed packet data, and utilized a flow-statistic-based mechanism to interpret metadata from the packet headers.
A detailed technical paper released by the researchers provides an in-depth exploration of the convolutional neural network (CNN) model developed for this purpose, featuring multiple layers and filters to enhance the accuracy of cyberattack detection.
Real-world tests conducted on the replica robot subjected to simulated cyberattacks against various systems yielded excellent results, with high identification accuracy, even after only 2-3 epochs of model training.
Optimized versions of this innovative protective system could find applications in similar, albeit more demanding, robotic scenarios, such as unmanned aircraft.
The researchers expressed their interest in examining the effectiveness of their intrusion detection system on different robotic platforms, like unmanned aerial vehicles, which operate at faster and more intricate dynamics compared to ground robots.
Do you want to donate to me?
No comments:
Post a Comment