Research On Accelerated Hardware Trojan Detection Through Transition Probability Analysis

Hardware Trojan is a tiny circuit which is maliciously implanted in an integrated circuit.It has the characteristics of small structure and serious damage.Hardware Trojans can be embedded in integrated circuits through electronic design automation(EDA)tools,the third-party intellectual property(IP)cores and foundry.Triggered under certain conditions the hardware Trojan would change circuit functions,leak circuit information,or stop the chip.Once the hardware Trojan is triggered,it would pose a serious threat to citizen privacy,information and communication,even national security.Therefore,it is especially important to guarantee the chips used in key areas have no hardware Trojans.Usually the hardware Trojan has a small structure and can be inserted in low activity nodes,so it's difficult to effectively detect whether the chip contains a hardware Trojan.In order to improve the successful detection efficiency of the hardware Trojan,this paper proposes a method of inserting 2-to-1 Multiplexers(MUXs)during the design process of integrated circuit to increase the transition probability of the nodes.This method inserts the MUXs at the main input of the nodes whose transition probability are lower than the threshold,which can increase the transition probability of the corresponding nodes.In this way,the hardware Trojan's activation probability is improved,and the efficiency of hardware Trojan detection is strikingly improved.The fan-out cone and the logic depth of the circuit are used to optimize the MUXs inserting algorithm to reduce the number of inserted MUXs,and decrease the circuit overhead caused by inserting the MUXs.By presetting the maximum delay ratio,the critical path delay is effectively controlled in an acceptable range,to prevent degradation of circuit performance caused by excessive delay of critical path.The experiment regards power consumption as side channel signal,and detects hardware Trojan by the side channel signal analysis method.The original circuit serves as a reference circuit for the control group and the original circuit inserted MUXs serves as a reference circuit for the experimental group.Design Compiler(DC)is utilized to perform logic synthesis of the circuits,and hardware Trojan is inserted into the netlists after synthesis as the test circuits of each group.Modelsim is the main tool to generate Value Change Dump(VCD)files for each group,and the PrimeTime-PX(PT-PX)is used for power simulation to obtain the circuits' power consumption information.The power consumption data is acquired by extracting the power consumption values at the same time of the reference circuit and the test circuit with the Tool Command Language(TCL)script.Using Matlab to perform Karhunen-Loeve(K-L)transformation for power consumption data and analyze the results.Taking ISCAS'89 benchmark circuit as the experimental object,the results show that after inserting MUXs the circuit' transition probability is significantly improved,the number of nodes whose transition probability are less than the threshold is greatly reduced,and the activation probability of the hardware Trojan inserted into the circuit is remarkably increased.The critical path delay is effectively controlled within the preset factor.Under a certain range of noise and area ratio of hardware Trojan,the results of K-L transformation show that before raising the transition probability,the power contribution value of hardware Trojan is small,and the deviation between the reference circuit and the testing circuit is small,so the projection results of the reference circuit and the testing circuit cannot be distinguished.And it can not effectively judge whether the test circuit contains a hardware Trojan.After raising the transition probability,it can increase the power contribution value of hardware Trojan to a certain extent.The projection results of the testing circuit are obviously different from the reference circuit.Features of the hardware Trojan are more noticeable and it can be effectively judged in the testing circuit.