基于随机森林的硬件木马检测方法

Hardware Trojans Detection Based on Random Forest

针对硬件木马特征多样性以及激活效率低的现状,提出了一种基于随机森林的硬件木马检测方法,即在门级电路检测触发节点。首先,从已知网表中提取每个节点的特征值;然后,根据时序电路和组合电路两种情况,通过随机森林分类器赋予每种特征相应的权重并生成两种模型(双模型用于识别未知网表中的可疑触发节点,并且给出每个可疑触发节点的可疑度结果);最后,通过可疑度排名的前n%个可疑节点检测硬件木马。实验结果证明了该方法的优越性。

In view of the diversity of hardware Trojans features and low activation efficiency,a hardware Trojans detection method based on random forest is proposed,which is to detect the trigger node in the gate level circuit.First,we extract the feature value of each node from the known netlist.Then according to the two cases of sequential circuit and combined circuit,the random forest classifier is used to give each feature a corresponding weight and generate two models,which are used to identify suspicious trigger nodes in the unknown netlist,and gives each suspicious trigger node suspicious results.Finally,hardware Trojans are detected by the top n%suspicious nodes ranked by suspiciousness.The experimental results shows the superiority.