Security of unmanned aerial vehicles: Dynamic state estimation under cyber-physical attacks

摘要

The goal of this work is to detect faults and cyber-physical attacks on unmanned aerial vehicles (UAVs) using dynamic state estimation to determine the nature of such vulnerabilities. We develop and introduce a distributed UAV architecture to characterize attacks and their propagation. Central to the distributed control architecture is a supervisory controller that relies on active sensing and actioning of control-lable contactors to collect the needed information for system state estimation using an adaptive algorithm. Uncertainty is reduced through adaptive reconfiguration of the system based on correlation of measurements from sensor readings and prior information on system state. We show that an adaptive greedy strategy algorithm guarantees theoretical worst-case performance for various cyber-physical attack scenarios. This is demonstrated by the results of prototype implementations and simulation on a subgraph of the UAV in the case of a gain control attack.