Unmanned Aerial System Security using Real-time Autopilot Software Analysis

摘要

Given the proliferation of Unmanned Aerial Systems (UASs) for military and commercial purposes, there is a fast-approaching need for comprehensive and robust systems which secure the UAS for public safety and assurance of military goals. This paper puts forth a comprehensive system of protecting UASs against hardware and software attacks, at the lowest and highest semantic levels. This approach monitors the autopilot directly, as well as the lowest level of the operating system on-board the UAS to establish whether or not malicious software has altered the function of the aircraft. Coupled with the other recent work by the authors, this completes a system that also monitors the lowest semantic level of events on the UAS through ob-board sensors, allowing for the detection of hardware failures as well as malicious attacks. After testing a quadcopter testbed with this anomaly detection system against software attacks, it can be seen that the system is successful in the detection of changes to the control flow of the autopilot, as well as the detection of any malicious events on the UAS's computer.