A failure physics model for hardware Trojan detection based on frequency spectrum analysis

Hardware Trojan embedded by adversaries has emerged as a serious security threat. Until now, there is no a universal method for effective and accurate detection. Since traditional analysis approaches sometime seem helpless when the Trojan area is extremely tiny, this paper will focus on the novel detection method based on frequency spectrum analysis. Meanwhile, a failure physics model is presented and depicted in detail. A digital CORDIC IP core is adopted as a golden circuit, while a counter is utilized as a Trojan circuit. The automatic test platform is set up with Xilinx FPGA, LabVIEW software, and high precision oscilloscope. The power trace of the core power supply in FPGA is monitored and saved for frequency spectrum analysis. Experimental results in time domain and frequency domain both accord with those of theoretical analysis, which verifies that the proposed failure physics model is accurate. In addition, due to immunity to vast measurement noise, the novel method processing in frequency domain is superior to the traditional method conducting in time domain. It can easily achieve about 0.1% Trojan detection sensitivity, which indicates that the novel detection method is effective.