学术文献库

Millimeter-wave radar systems are one of the core components of the safety-critical Advanced Driver Assistant System (ADAS) of a modern vehicle. Due to their ability to operate efficiently despite bad weather conditions and poor visibility, they are often the only reliable sensor a car has to detect and evaluate potential dangers in the surrounding environment. In this paper, we propose several attacks against automotive radars for the purposes of assessing their reliability in real-world scenarios. Using COTS hardware, we are able to successfully interfere with automotive-grade FMCW radars operating in the commonly used 77GHz frequency band, deployed in real-world, truly wireless environments. Our strongest type of interference is able to trick the victim into detecting virtual (moving) objects. We also extend this attack with a novel method that leverages noise to remove real-world objects, thus complementing the aforementioned object spoofing attack. We evaluate the viability of our attacks in two ways. First, we establish a baseline by implementing and evaluating an unrealistically powerful adversary which requires synchronization to the victim in a limited setup that uses wire-based chirp synchronization. Later, we implement, for the first time, a truly wireless attack that evaluates a weaker but realistic adversary which is non-synchronized and does not require any adjustment feedback from the victim. Finally, we provide theoretical fundamentals for our findings, and discuss the efficiency of potential countermeasures against the proposed attacks. We plan to release our software as open-source.