Beyond Detection: Securing Robotic Vehicles

Abstract

Autonomous Robotic Vehicles (RV) such as drones and rovers rely exclusively on sensors to perceive their physical states, and plan their trajectory using specialized algorithms. Physical attacks such as sensor tampering and spoofing can feed erroneous sensor measurements through physical channels that may disrupt RVs, and result in mission failures. Prior work in this area is largely limited to attack detection. However, detection alone is not a complete solution, as it does not ensure safety. We propose a novel control framework to mitigate physical attacks against RVs, and ensure safety and security. Our framework allows RVs to operate safely despite malicious attacks, can easily integrate with the RV’s control loop and trajectory planning algorithms, and do not require any changes to the hardware.