A PCA and Mahalanobis distance-based detection method for logical hardware Trojan

Due to wide application of FPGA devices, which are easy to be attacked by HTs, it has been risen up more concern on chip security. In order to ensure the efficiency of detection, side-channel analysis is used to detect HTs. However, existing methods for HT detection cannot detect the LHTs with the small size, which takes up lower than 1% area of the whole chip, considering that signal of LHTs is drowned in process deviations. To complete the side-channel analysis for LHT, we proposed an algorithm that combines PCA and Mahalanobis distance to improve the accuracy of LHTs detection. In addition, performed on FPGA of Xilinx, a sample of LHTs was designed and verified, which took up nearly 0.1% area of the whole chip. The influence of process deviation on the power consumption of FPGA was simulated by Hspice. A set of power-frequency data of golden FPGA was generated by Matlab as panel data, which was considering the influence. Another two sets of FPGA with and without LHT were generated to test the accuracy of the proposed method. For LHT and golden FPGA, the detection accuracy of proposed method is 99% and 87%, respectively.