Neural Networks & Logistic Regression for FPGA Hardware Trojan Detection

A hardware Trojan is a malware of increasing importance due to the increase in the growing number of digital circuits. A hardware Trojan can enter the circuit at any stage of chip manufacturing. The first step to deal with this malware is to detect the presence of this withering factor inside an integrated circuit. Since the appearance of this malware, various methods including thermal image processing have been proposed to detect hardware Trojans. The first challenge for research in this field is the lack of a database available for thermal images of Trojan chips. Accordingly, different images were taken from a Trojan-affected FPGA using a T4 thermal camera in this study. Our database includes 12 series of thermal images which are captured for each chip 55 seconds after programming. Then, two different methods have been proposed to detect hardware Trojans in the created database. In this paper, we propose a thermal image processing-based Hardware Trojan detection method on FPGA chips using neural networks, assuming the availability of a golden chip. Results demonstrate that if our method is combined with the previous method, the detection rate can be increased significantly.