Generating Adversarial Examples for Hardware-Trojan Detection at Gate-Level Netlists

Recently, the great demand for integrated circuits (ICs) drives third parties to be involved in IC design and manufacturing steps. At same time, threat of injecting a malicious circuit, called hardware Trojan, by has been increasing. Machine learning is one powerful solutions detecting Trojans. However, weakness such machine-learning-based classification method against adversarial examples (AEs) reported, which causes misclassification adding perturbation input samples. This paper firstly proposes framework generating hardware-Trojan detection at gate-level netlists utilizing neural networks. The proposed replaces Trojan with logically equivalent ones, makes it difficult detect them. Secondly, we propose Trojan-net concealment degree (TCD) modification evaluating value (MEV) as measures amount modifications. Finally, based on MEV, pick up patterns apply detection. experimental results using benchmarks demonstrate that successfully decreases true positive rate (TPR) maximum 30.15 points.