Hardware Trojan Detection Based on Cluster Analysis of Mahalanobis Distance

摘要

As the threats of hardware Trojan becoming increasingly serious, we proposed a new hardware Trojan detection method based on class center Markov distance classification in this paper. After modeling the power characteristics of hardware Trojan, we demonstrated the feasibility of using class center distance to detect the hardware Trojan, and since the shortcomings of Euclidean distance in this area, we finally selected the Mahalanobis distance as the detection basis. Compared with the traditional hardware Trojan detection method, the process of our method is targeted optimized, and the detection accuracy is also improved. Experimental results show that, compared with the hardware Trojan detection method based on Euclidean distance, even on the case that the Trojan scale is small our method can still achieve a higher detection success rate.