Introduction

To improve road safety and driving experiences, autonomous vehicles have emerged recently, and they can sense their surroundings and navigate without human intervention. Although promising and improving safety features, the trustworthiness of these cars has to be examined before they can be widely adopted on the road. Unlike traditional network security, autonomous vehicles rely heavily on their sensory ability of their surroundings to make driving decision, which makes sensors an interface for attacks. Thus, in this project we examine the security of the sensors of autonomous vehicles, and investigate the trustworthiness of the ‘eyes’ of the cars.

Our work investigates sensors whose measurements are used to guide driving, i.e., millimeter-wave radars, ultrasonic sensors, forward-looking cameras. In particular, we present contactless attacks on these sensors and show our results collected both in the laboratory and outdoors on a Tesla Model S automobile. We show that using off-the-shelf hardware, we are able to perform jamming and spoofing attacks, which caused the Tesla’s blindness and malfunction, all of which could potentially lead to crashes and impair the safety of self-driving vehicles.