Attacking and defending multi-robot systems

Abstract

Multi-robot systems have offered a solution to completing tasks efficiently. However, with such large systems also comes a large attack surface, presenting itself as a possible security risk in the cyber world. This project aims to explore Byzantine attacks on the navigation and exploration workloads in ROS-based multi-robot systems. By analysing the architecture of these workloads and their implementations, several attacks were identified with the aim of disrupting each workload execution, either through delay of the workload completion or causing the workload to fail to accomplish its goals. Through simulating the attacks on the workloads with a ROS-based simulator, the report evaluates the severity of each attack to show the vulnerabilities in the existing implementations of these ROS-based systems. Additionally, several defensive controls were proposed to improve the robustness of these workload implementations.