摘要： 针对目前汽车CAN总线入侵检测算法性能模糊测试方法因测试用例覆盖率低、针对性差而导致的测试结果可信度不高的问题，提出一种改进的汽车CAN总线入侵检测算法性能模糊测试方法。针对是否已知CAN总线协议规范的情况分别基于字段权重和改进Wasserstein生成对抗网络(WGAN-GP)生成模糊测试用例，对KNN算法和AdaBoost算法进行了测试，测试结果表明，AdaBoost算法的检测性能优于KNN算法。试验验证了所提出的测试方法用于测试入侵检测算法的性能能够得到可信度较高的试验结果，达到了为汽车CAN总线入侵检测算法的选用提供参考依据的目的。

中图分类号： TP306.2
文献标识码： A
DOI： 10.19358/j.issn.2096-5133.2022.04.005
引用格式： 田韵嵩，李中伟，谭凯，等. 汽车CAN总线入侵检测算法性能模糊测试方法研究[J].信息技术与网络安全，2022，41(4)：32-38.

Research on fuzzy test method of the detection ability of in-vehicle CAN bus intrusion detection algorithm

Abstract： The test results of the current vehicle CAN bus intrusion detection algorithm performance fuzzy test method are not highly reliable, due to the low test case coverage and poor pertinence. Aiming at this problem，an improved in-vehicle CAN bus intrusion detection algorithm performance fuzzy test method was proposed. According to whether the CAN bus protocol specification was known or not, fuzzy test cases were generated based on field weights or improved Wasserstein Generative Adversarial Network(WGAN-GP). The generated test cases were used to test the KNN algorithm and the AdaBoost algorithm. The test results showed that the detection performance of the AdaBoost algorithm was better than that of the KNN algorithm. The test verified that the test method proposed in this paper can obtain the test results with high reliability when used to test the performance of the intrusion detection algorithm, and achieved the purpose of providing a reference for the selection of the intrusion detection algorithm of the in-vehicle CAN bus.

Key words : intrusion detection algorithm；detecting ability test；Controller Area Network(CAN)；Generative Adversarial Network(GAN)； fuzzy test