Poster

Abstract

Robotic vehicles are playing an increasingly important role in our daily life. Unfortunately, attackers have demonstrated various sensor spoofing attacks that interfere with robotic vehicle operations, imposing serious threats. Thus, it is crucial to discover such attacks earlier than attackers so that developers can secure the vehicles. In this paper, we propose a new sensor fuzzing framework SensorFuzz that can systematically discover potential sensor spoofing attacks on robotic vehicles. It generates malicious sensor inputs by formally modeling the existing sensor attacks and leveraging high-fidelity vehicle simulation, and then analyzes the impact of the inputs on the vehicle with a resilience-based feedback mechanism.