"Automotive Binary Programs"

"First, 30 automotive parts systems from different vehicles are selected, and the data is parsed to obtain the file systems of various automotive components. To protect the interests of automotive companies and reduce legal disputes, the names of the enterprises involved in the benchmark will not be disclosed. In this study, we selected parts systems from 30 vehicle models across 8 companies, comprising 20 new energy vehicles and 10 fuel vehicles.The 30 parts systems are subjected to data parsing to obtain corresponding embedded file system data. Binary programs from directories such as \/bin, \/lib, \/lib64, \/system\/bin, \/system\/lib, and \/system\/lib64in these embedded file systems are extracted to form a candidate dataset.In the candidate dataset, files are categorized and counted based on the first part of their names to determine the frequency of their occurrences. Since data from four shared library directories (\/lib, \/lib64, \/system\/lib, and \/system\/lib64) are selected from each parts system, programs with a file frequency of less than 4 are removed. The remaining programs are used in at least 2 vehicle models, indicating cross-model applicability.Subsequently, expert manual analysis is conducted on the processed data to identify risks present in different programs. We selected 216 programs containing specific vulnerabilities (positive samples) and 4 programs without specific vulnerabilities (negative samples) as the benchmark."