基于多信号特征融合的硬件木马识别技术  

Hardware Trojan detection based on combination of multiple features

作　　者：赵聪慧 严迎建[1] 刘燕江 朱春生 ZHAO Cong-hui;YAN Ying-jian;LIU Yan-jiang;ZHU Chun-sheng(Key Laboratory of Information Security,Information Engineering University,Zhengzhou 450000,China)

机构地区：[1]信息工程大学信息安全重点实验室,河南郑州450000

出　　处：《计算机工程与设计》2021年第12期3365-3372,共8页Computer Engineering and Design

摘　　要：针对现有基于信号特征的硬件木马检测方法中存在木马特征集单一、检测精度低和普适性差等问题,提出一种基于多信号特征融合的硬件木马识别方法。通过分析硬件木马的隐藏性,建立触发节点植入与载荷节点植入的硬件木马隐藏性模型,构造低静态翻转率、低动态翻转率、低组合0可控性、低组合1可控性和低组合可观察性的硬件木马特征集,利用KNN算法建立硬件木马检测模型。实验结果表明,该方法达到了98.23%的木马信号平均识别率,与文献[3]和文献[15]相比,分别提高了16.30%和10.24%,大幅提升了木马检测能力。In the existing signal feature-based hardware Trojan detection methods,there are problems of single Trojan feature set,low detection accuracy and poor universality.Therefore,a hardware Trojan detection utilizing multiple dimensional signal features was proposed.By analyzing the concealment of hardware Trojan,a concealment model for trigger node implantation and load node implantation was established.A hardware Trojan detection model was established using the KNN algorithm,with the feature set of low static flip rate,low dynamic flip rate,low combination 0 controllability,low combination 1 controllability and low combination observability.Experimental results show that the average recognition rate of Trojan signals reaches 98.23%,which is 16.30%and 10.24%higher than literature[3]and literature[15],greatly improving the Trojan detection ability.

关 键 词：硬件木马检测 门级网表 信号特征 隐藏性 机器学习 KNN分类算法 

分 类 号：TP309.1[自动化与计算机技术—计算机系统结构]