Hardware Trojan (HT) is a crucial problem in the integrated circuits and digital systems industry, which has many malicious impacts on circuits such as overheating, functionality changes, and information leakage. Although many detection methods have been studied to detect HTs in circuits, a comprehensive vulnerability analysis of digital circuits to HTs has not been investigated yet. This paper presents a novel digital circuits vulnerability analysis to HTs using convolutional neural networks (CNN) with high accuracy. First, we scrutinize vulnerable factors in circuits, and we generate vulnerability heat maps in circuits’ regions. Then, an appropriate neural network architecture is designed. Finally, we employ neural networks to classify different regions vulnerability level. With our approach, vulnerable regions of digital circuits can be classified with 92.5% average accuracy.
