基于K-Means聚类和集成学习的HTD仿真

Hardware Trojan Detection Simulation Based on K-Means Clustering and Ensemble Learning

为了实时检测信息系统中是否存在硬件木马,提出基于K-Means聚类和集成学习的硬件木马检测方法。采用基于信息熵改进的K-means动态聚类算法去除冗余数据,挖掘信息系统硬件运行的有效数据;在该数据中采用基于改进旋转森林的集成学习方法建立识别硬件木马的文本分类器,引入动态加权投票集成方法,检测出硬件木马。仿真结果显示,所提方法硬件木马数据检测率高达99%,误报率最大值仅为3%;可以实时检测出硬件木马,不存在时延。和同类检测方法相比,所提方法对硬件木马的检测精度、检测实时性存在优越性。

In order to detect hardware Trojan in information system in time, this paper proposes a detection method of hardware Trojan based on K-means clustering and ensemble learning.Based on the improved k-means dynamic clustering algorithm of information entropy, the redundant data were deleted, and the effective data of information system hardware operation were mined.In this effective data, according to the improved ensemble learning method of revolving forest, the text classifier of hardware Trojan was founded.The integration method of dynamic weighted voting was introduced to detect hardware Trojan.The simulation results show that this method has high detection rate(99%) and low false alarm rate(3%), being better than other similar detection methods.