基于PSO-SVM的硬件木马检测  

HARDWARE TROJANS DETECTION BASED ON PSO-SVM

作　　者：李莉云 伍忠东[1] 芦德钊 Li Liyun;Wu Zhongdong;Lu Dezhao(School of Electronic and Information Engineering,Lanzhou Jiaotong University,Lanzhou 730070,Gansu,China)

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《计算机应用与软件》2023年第6期310-314,336,共6页Computer Applications and Software

基　　金：甘肃省高等学校创新团队项目(2017C-09);兰州市科技局科技项目(2018-1-51)。

摘　　要：针对基于门级网表的硬件木马识别方法中硬件木马识别率低且识别效果不稳定的问题,提出一种基于粒子群优化支持向量机(Particle Swarm Optimization-Support Vector Machine,POS-SVM)算法的硬件木马检测方法。对电路的门级网表特征分析提取出7维特征,利用SMOTE算法对数据进行预处理,改善数据集类别不平衡问题,使用该特征训练支持向量机分类器,用粒子群算法对支持向量机的参数优化以获得较高的识别准确率,利用该分类器达到识别硬件木马的目的。实验结果表明,该算法提高了硬件木马的识别率,实现99.47%线网识别准确率。Aimed at the problem of low recognition rate and unstable recognition effect of hardware Trojans classification method based on gate-level netlists,a hardware Trojans detection method based on particle swarm optimization-support vector machine(POS-SVM)algorithm is proposed.The gate-level netlist features of the circuit were analyzed to extract 7-dimensional features.The synthetic minority oversampling technique(SMOTE)algorithm was used to preprocess the data set to overcome the imbalance of the feature data set.This data set was used to train the SVM classifier.PSO was used to optimize the parameters of SVM to obtain a higher recognition accuracy.The classifier was used to achieve the purpose of identifying the hardware Trojans.The experimental results show that the algorithm improves the recognition rate of the hardware Trojans,and the network recognition accuracy achieves 99.47%.

关 键 词：硬件木马 门级网表 支持向量机 SMOTE算法 粒子群算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]