Research On Hardware Trojan Detection Based On Path Delay

As the cornerstone of an information society,integrated circuits occupies a vital position in social life and national security.Under the influence of economic globalization,the semiconductor industry outsources design and fabrication phase,which makes integrated circuits more vulnerable to Hardware Trojans at design and fabrication time,thus causing serious harm to chip security.Therefore,the research on Hardware Trojan detection technology is of great significance to the security of the chip.In the traditional Hardware Trojan detection method based on the side-channel delay,the extra delay changes caused by the process deviation will cover up the influence of the Hardware Trojan.Aiming at the above drawback,a Hardware Trojan detection method based on relative delay fingerprint is proposed.Based on the detection method a complete detection process is established.In the proposed method,the shortest paths including vulnerable nets is collected to construct the suspicious path set at the gate-level stage of the circuit,and then the relative delay fingerprint is constructed using the delay difference between the paths.the relative delay data of the circuits under test is compared with the generated fingerprint to judge whether there is a Hardware Trojan.Firstly,aiming at the influence of process deviation on path delay,Monte Carlo simulation is performed on the path delay using HSPICE.The results show that under the same process deviation,the delay difference between paths is much smaller than the change in path delay,thus indirectly verifying the advantages of the relative delay fingerprint.Afterwards,the lowactivity nets and low-testability nets of the circuit are obtained through random test vector simulation and SCOAP algorithm analysis,so as to obtain the suspicious path of the Hardware Trojan in the circuit.Finally,the clock-sweeping method is utilized to map the Hardware Trojan into a delay fault to complete the measurement of the path delayIn order to verify the feasibility and effectiveness of the proposed method,the s38584 circuit in the ISCAS'89 benchmark circuits is selected as the carrier circuit,and the search for vulnerable nets and suspicious paths is implemented using C language.The backend design and transition-delay fault test vectors are then performed on the genuine circuits.Finally,the process variation library is simulated in Perl.After simulating the circuits with the test vector,the delay data is collected and the relative delay data is constructed using C language to complete the generation of the relative delay fingerprint.In order to verify the effectiveness of the detection method,the realative delay data of the four haraware Trojan circuits are extracted respectively.Due to the high dimensionality of the generated data,the MDS dimensionality reduction algorithm is implemented to reduce the dimensionality of the relative path data,and the reduced dimensionality of the relative delay fingerprint is constructed into a golden convex hull.The existence of Hardware Trojan is judged by comparing the distance between the data point of the circuit under test and the golden convex hull.In order to show the advantages of relative delay fingerprints,both the traditional delay fingerprint and relative delay fingerprint detection experiments are performed on four Trojan circuits.The results show that compared with the traditional side-channel delay detection method,using the proposed method can well reduce the additional delay changes caused by process deviations and achieve better Hardware Trojan detection effect.In addition,there is no additional area consumption brought to the circuit,and under the conditions of process deviation simulation,there is still a good detection effect for the Hardware Trojans with an area ratio of 0.0385% in the proposed method.