Scholarlink

<p>This paper presents a lightweight, hierarchical Intrusion Detection System based on a dilated convolutional neural network (HDCNN) specifically designed for IoT networks. A key innovation in this paper is the integration of the weighted grey wolf optimization (WGWO) algorithm for feature selection. WGWO enhances the traditional GWO by better mimicking the hunting behavior of grey wolves, effectively identifying the most relevant features, reducing dataset dimensionality, and improving the efficiency of the lightweight HDCNN. HDCNN framework processes network traffic through three levels: first, second, and third. Higher levels identify broad patterns to distinguish normal from abnormal behavior, while third level analyze fine-grained packet details, such as attack types. The middle layer refines abnormal data analysis by cross-verifying additional information. This structured approach enables efficient data processing, balancing detail and abstraction without excessive computational demands. The evaluation has been done using real-world IoT datasets, namely BoT-IoT, Edge-IIoT, and CICIoT2023. The results show that WGWO effectively reduced the number of selected features to 6, 10, and 7 for BoT-IoT, Edge-IIoT, and CICIoT2023, respectively. In terms of computational efficiency, WGWO outperformed traditional GWO in both average runtime and the number of iterations required. Furthermore, HDCNN outperforms other related approaches in the literature in terms of accuracy and F-score. It attained detection accuracies of 100, 99.78, and 99.86% for BoT-IoT, Edge-IIoT, and CICIoT2023, respectively.</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

Hierarchical lightweight intrusion detection system using deep learning in the context of IoT

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Orieb Abualghanam,
  • Hadeel Alazzam,
  • Wesam Almobaideen

摘要

This paper presents a lightweight, hierarchical Intrusion Detection System based on a dilated convolutional neural network (HDCNN) specifically designed for IoT networks. A key innovation in this paper is the integration of the weighted grey wolf optimization (WGWO) algorithm for feature selection. WGWO enhances the traditional GWO by better mimicking the hunting behavior of grey wolves, effectively identifying the most relevant features, reducing dataset dimensionality, and improving the efficiency of the lightweight HDCNN. HDCNN framework processes network traffic through three levels: first, second, and third. Higher levels identify broad patterns to distinguish normal from abnormal behavior, while third level analyze fine-grained packet details, such as attack types. The middle layer refines abnormal data analysis by cross-verifying additional information. This structured approach enables efficient data processing, balancing detail and abstraction without excessive computational demands. The evaluation has been done using real-world IoT datasets, namely BoT-IoT, Edge-IIoT, and CICIoT2023. The results show that WGWO effectively reduced the number of selected features to 6, 10, and 7 for BoT-IoT, Edge-IIoT, and CICIoT2023, respectively. In terms of computational efficiency, WGWO outperformed traditional GWO in both average runtime and the number of iterations required. Furthermore, HDCNN outperforms other related approaches in the literature in terms of accuracy and F-score. It attained detection accuracies of 100, 99.78, and 99.86% for BoT-IoT, Edge-IIoT, and CICIoT2023, respectively.