Controller area network (CAN) bus system is widely accepted for connecting Electronic Control Units (ECUs) on in-vehicle networks. Fuzz-testing (i.e., testing with anomalous input data, often pseudo-random) over the CAN-bus is considered a proven method to detect security vulnerabilities within the in-vehiclenetwork. The automatic execution of the test cases and checking the robustness makes CAN-bus fuzzing a popular choice in the automotive testing community. However, due to the large number of fuzz testcases, the execution logs are often large for CAN-bus fuzzing. Root cause analysis of these logs is not a trivial task and requires expert manual effort as not all the failures are security critical and some failures can be contextual issues or false positives. Thus, automatic determination of the class, relative severity, and robustness of a potential failure is desirable for prompt analysis of the vulnerabilities. In this workwe propose a testing architecture to automate the determination of the potential security vulnerabilities from the vast fuzz testing logs and generate a risk score associated with the failure class, relative severity, and robustness of the failure. We also identify different classes of the vulnerabilities can be found in the in-vehicle networks, with the techniques to measure the relative severity, and robustness of the failures. In addition, we provide direction to implement a scoring system to quantify the risk associated with a potential vulnerability.
Kayas et al. (2023) studied this question.