Attack Detecting for the Multi-robots Systems: An Integrated Cyber-Physical Approach

To address security issues and effectively detect cyber-attack behaviors in multi-robot systems, this paper proposes a cyber-attack detection method that combines information-physical features. The method involves performing cyber intrusions, such as false data injection attack, replay attack, and DoS attack, on a multi-robot system using the Kali Linux operating system to disrupt its collaborative motion process. Physical state information of the system and network traffic information are then collected. Based on Convolutional Neural Network (CNN), a cyber-attack detection method is designed that fuses the information-physical features. The proposed method is validated on a multi-robot system in the Gazebo simulation environment. Simulation and experimental results demonstrate that the proposed cyber-attack detection method can effectively detect system anomalies with higher accuracy compared to the detection methods based solely on physical features or network traffic characteristics.