Investigating Malware-in-the-Loop Autopilot Attack Using Falsification of Sensor Data

Abstract

The proliferation of unmanned air systems (UAS) in recent years has provided advanced capabilities in many fields and industries. However, this widespread usage of UAS has also introduced many cyber threats. One of the primary threats faced by UAS is the threat of hijacking by cyber attack. This paper seeks to investigate and understand this threat by implementing and validating a cyber hijacking attack on a UAS using a software-in-the-loop simulation. The relevant control theory necessary to develop the attack is explained and demonstrated on a generic linear second-order system, and the attack is implemented in a Gazebo simulation and demonstrated on a quadcopter using the ArduCopter autopilot. This simulation permitted evaluation of the attack throughout several experiments. The experiments allowed for evaluation of the attack under multiple different flight conditions. Simulation results presented in the paper demonstrate that the attack poses a legitimate threat to UAS. The attacker demonstrates successful performance when the vehicle is attempting to hold position, but demonstrates only marginal control over the vehicle when it is attempting to fly to a desired position.