基于改进DCGAN的车载CAN总线模糊测试方法  

Fuzzing Testing Method for Vehicle CAN Bus Based on Improved DCGAN

作　　者：周云鹏 张爱梅[1] ZHOU Yun-peng;ZHANG Ai-mei(School of Mechanical and Power Engineering,Zhengzhou University,Zhengzhou Henan 450000,China)

机构地区：[1]郑州大学机械与动力工程学院,河南郑州450000

出　　处：《计算机仿真》2023年第10期176-180,186,共6页Computer Simulation

摘　　要：为实现无需协议的先验知识进行车载网络安全漏洞检测,提出一种基于深度卷积生成对抗网络的测试用例生成方法。通过引入卷积注意力机制来改进DCGAN,从空间和通道子模块出发,提高模型的特征表现力和生成CAN报文的格式正确性。搭建合适的生成对抗网络模型,将预处理后的CAN报文输入到模型中生成测试用例,对目标协议进行模糊测试。实验结果表明,所提方法能够有效发现总线协议的安全问题,且改进后的模型具有更高的测试用例通过率(87.7%)和漏洞发现能力(4.81%),测试效率显著提高。In order to achieve vehicle network security vulnerability detection without prior knowledge of the protocol,a test case generation method based on deep convolutional generative adversarial network was proposed.The DCGAN was improved by introducing a Convolutional Block Attention Module,started from the spatial and channel sub-modules,to improve the feature expressiveness of the model and the format correctness of the generated CAN message.Constructed a suitable generative adversarial net model,input the preprocessed CAN message into the model to generate test cases,and fuzzed the target protocol.Experimental results show that the proposed method can effectively found the security problems of the bus protocol,and the improved model has a higher test case pass rate(87.7%)and vulnerability discovery ability(4.81%),and the test efficiency is significantly improved.

关 键 词：车载安全 模糊测试 生成对抗网络 测试用例 卷积注意力机制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]