MLDet:基于结构特征和XGBoost的硬件木马检测方法  

MLDET: HARDWARE TROJAN DETECTION METHOD BASED ONSTRUCTURAL FEATURES AND XGBOOST

作　　者：杨欢 李海明 Yang Huan;Li Haiming(School of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 201306,China)

机构地区：[1]上海电力大学计算机科学与技术学院,上海201306

出　　处：《计算机应用与软件》2023年第11期302-307,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61772327)。

摘　　要：随着集成电路(Integrated Circuit,IC)的规模日益庞大,IC的生产和制造全球化,不受信任的第三方很容易将硬件木马插入知识产权内核,从而威胁IC的安全。因此,有必要研究硬件木马的检测方法,特别是IC设计阶段的硬件木马检测。提出一种名为MLDet的方法,提取门级网表结构特征;用XGBoost算法来检测硬件木马。MLDet从已知网表中提取木马特征值,并使用XGBoost算法训练;将训练好的检测模型用于未知网表的检测;成功将网表中的节点分类为普通节点和木马节点。实验结果表明,MLDet获得了85.60%的平均硬件木马检测率,部分基准电路的平均硬件木马检测率达到100%。With the increasing scale of integrated circuit(IC)and the globalization of IC production and manufacture,it is easy for the untrusted third party to insert hardware trojans into the intellectual property core,thus threatening the security of IC.Therefore,it is necessary to develop the detection methods of hardware Trojan,especially hardware Trojan detection in the IC design stage.A method called MLDet is proposed.It extracted the structural features of the gate-level netlist,and used the XGBoost algorithm to detect the hardware Trojans.MLDet extracts the Trojans feature values from the known Netlist and trains them by using XGBoost algorithm.The trained detection model was applied to the detection of the unknown Netlist.The nets in the Netlist were successfully classified as ordinary nets or Trojan nets.The experimental results show that MLDet can achieve an average hardware Trojan detection rate of up to 85.60%,and the hardware Trojan detection rate of some benchmark circuits reaches 100%.

关 键 词：硬件木马 机器学习 门级网表 特征提取 节点分类 

分 类 号：TN432[电子电信—微电子学与固体电子学] TP309.1[自动化与计算机技术—计算机系统结构]