Synergizing GCN and GAT for Hardware Trojan Detection and Localization

Yu-Chen Hsiao, Chia-Heng Yen, Bo-Yang Ke, Kai-Chiang Wu

Abstract: Hardware Trojan (HT) is a common issue for the outsourcing model and it poses various threats to hardware security. HT may be implanted during the design phase through the use of open-source resources and uncertified tools. In this paper, we propose a novel synergistic graph convolutional network and graph attention network (SGCAT)-based method for HT detection in pre-layout register-transfer level (RTL) designs. The proposed method combines the strengths of graph convolutional neural network (GCN) and graph attention network (GAT) to provide the precise detection and localization of HTs in RTL designs. From the observation of the experimental results, the proposed method demonstrates better performance in terms of accuracy, F1-score, precision and recall for HT detection.