Scholarlink

<p>Hardware Trojan (HT) emerged as a serious security threat in the integrated circuit (IC) industry due to the globalization of the semiconductor supply chain. In recent days, with the exponential growth of computing power, different machine learning (ML) approaches enter into the field of hardware Trojan detection with more accuracy. Using ML algorithms, the features of HT are extracted from complex IC design and used for further processing to gain good accuracy. This research focuses on detecting hardware Trojans using an artificial neural network based on skip connections method. The proposed approach is based on ISCAS Trust-hub benchmark circuits using various classification metrics. At first, the Trojan features are extracted from gate level netlists. A multi-layer perceptron (MLP) with skip connections is then trained to identify Trojans using a newly constructed feature set derived from 56 Trojan net features. The experimental findings demonstrate that the suggested approach outperforms the prior state-of-the-art methods, achieving an average accuracy of 99.96%, average precision of 94.33%, and average F-measure of 91.08%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A New Hardware Trojan Detection Approach Using Skip Connection Based Multilayer Perceptron Network

  • 期刊论文

  • 发表日期：

    2026年4月4日

查看全文
  • Banalata Bhunia,
  • Malay Kule,
  • Hafizur Rahaman

摘要

Hardware Trojan (HT) emerged as a serious security threat in the integrated circuit (IC) industry due to the globalization of the semiconductor supply chain. In recent days, with the exponential growth of computing power, different machine learning (ML) approaches enter into the field of hardware Trojan detection with more accuracy. Using ML algorithms, the features of HT are extracted from complex IC design and used for further processing to gain good accuracy. This research focuses on detecting hardware Trojans using an artificial neural network based on skip connections method. The proposed approach is based on ISCAS Trust-hub benchmark circuits using various classification metrics. At first, the Trojan features are extracted from gate level netlists. A multi-layer perceptron (MLP) with skip connections is then trained to identify Trojans using a newly constructed feature set derived from 56 Trojan net features. The experimental findings demonstrate that the suggested approach outperforms the prior state-of-the-art methods, achieving an average accuracy of 99.96%, average precision of 94.33%, and average F-measure of 91.08%.