SEBD: Sensor Emulation Based Backdoor for Autopilot

Abstract

Autopilot is an important component of robotic vehicles. It comprises the hardware and software utilized to automatically control and drive robotic vehicles. Due to its extensive usage and significant role in robotic vehicles, its security has garnered considerable attention. Currently, there are many open-source autopilot hardware options produced by vendors and distributed through intermediaries. These open-source autopilot hardware solutions are utilized by companies and individuals to build their own robotic vehicles. However, in the supply chain, there is potential risk that the autopilot hardware might be modified. Due to the unwavering trust in open source solutions and limited familiarity with autopilot hardware, developers of robotic vehicles might inadvertently procure hardware that has been deliberately modified by attackers. In this research, we proposed a stealthy, low-cost, hardware based supply chain attack that targets autopilot hardware. This attack is independent of the autopilot software, it is malicious to any normal autopilot software that is executed by the infected hardware.