Research On Hardware Trojan Intelligent Diagnostic Method Based On Gate-level Netlist

With the development of information technology in the world,integrated circuit(IC)has become an indispensable and important part of modern society.In order to reduce the production cost of integrated circuit as much as possible,the design,manufacturing,and assembly stages of integrated circuit are contracted by different third-party factories,which provides an advantage for attackers to implant hardware Trojan in integrated circuits.The security of integrated circuits has attracted a lot of attention,but there is no good tool for locating and removing hardware Trojan in circuit design.The dissertation takes the gate-level netlist as the research object,detects whether there is a hardware Trojan in the netlist,and deletes the hardware Trojan in the netlist according to the topology of the netlist.The main research work of this dissertation is as follows:(1)The basic concept and structure of the hardware Trojan are introduced,the performance form of the hardware Trojan at the gate-level netlist surface layer is expounded,the properties and classification of the hardware Trojan are summarized,several main hardware Trojan detection methods are analyzed and their advantages and disadvantages are compared.(2)An unsupervised hardware Trojan detection algorithm based on gate-level netlist is studied.For hardware Trojan in gate-level netlists,an unsupervised clustering method is designed to distinguish the Trojan netlists from the ordinary netlists.The method includes four steps: constructing a directed graph model,feature extraction,net clustering,and category judgment.Using this method,suspicious net set in the Trojan netlist can be extracted.The directed graph model is used to quickly access net objects.Net clustering uses K-Means and DBSCAN clustering algorithms.We use experiments to verify the Trojan detection effect of this method,the results show that this method can distinguish Trojan netlist and ordinary netlist 100%.The Trojan detection results of net clustering algorithms K-Means and DBSCAN are equivalent,but K-Means has a small time overhead.(3)A hardware Trojan diagnostic method based on suspicious net set is designed.For the suspicious net set extracted from the above hardware Trojan detection method,a method to completely eliminate the hardware Trojan in the gate-level netlist is designed.This method first maps the discrete nets into logical units,then uses the netlist topology to locate the hardware Trojan,and finally removes the redundant entries of the logical unit according to the SCOAP controllability value to eliminate the hardware Trojan in the gate-level netlist.Finally,the test benchmark in Trust-Hub is used to verify the correctness of this method.Experimental results show that this method can completely delete the Trojan structure without affecting the normal structure of the netlist.