Research and Implementation of Fuzzy Test Method Based on Multi-armed Bandit
Yiwen Qin, Jincheng Jiang, Changzhi Li, Futao Lu, Kaiwen Deng, Zengyu Yang · 2024
Thereare a lot of vulnerabilities in industrial control protocols due to rothole in protocol implementation and problems with exception handling of function codes. In order to solve low effectiveness of generating test cases and the inability to adjust the test case generation by feedback from the device, fuzzy test method based on multi-armed bandit(MAB) algorithm is proposed. Firstly, the protocol unit is divided into feature sets and a matrix model is established by deeply analyzing the protocol specification, so as to initially preprocess the input data set. Secondly, the protocol constraint rule set is constructed to filter the invalid cases that violate the protocol specification. Finally, the exploration and utilization phase of the model is customized by using multi-armed bandit algorithm and fuzzy test method to increase the selection probability of mutation factors that easily trigger vulnerabilities, and establish a test case queue to save the fuzzy test cases, moreover, the probability distribution is adjusted by using the feedback of test results.Experimental results show that this method improves the abnormality rate of the test system, and the vulnerability mining ability is significantly improved.