Scholarlink

<p>The Internet of Things (IoT) has greatly improved connectivity and automation, but its rapid growth has introduced significant security vulnerabilities. To tackle these issues, this study proposes an Intrusion Detection System (IDS) tailored for IoT environments. The proposed model incorporates a Radial Basis Function Neural Network (RBFNN) for precise classification, the Whale Optimization Algorithm (WOA) for feature selection, and an Autoencoder for robust outlier detection. The model was trained and assessed using the NF-ToN-IoT and NF-Bot-IoT datasets, creating a combined dataset (NF-IoT) and an upgraded version (NF-IoT-v2) featuring a two-layer outlier detection strategy. The experimental results show that the proposed IDS achieves 97.25% accuracy (ACC) and 85.77% Matthews Correlation Coefficient (MCC), demonstrating its efficacy in both anomaly detection and classification. These findings underscore the model’s effectiveness in securing IoT networks against cyber threats. Future improvements include enhancing the feature selection method and investigating alternative outlier detection strategies to further optimize model performance.</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

Optimizing abnormal activities in IoT networks using IDS with deep learning and feature engineering

  • 期刊论文

  • 发表日期：

    2025年9月11日

查看全文
  • Mouaad Mohy-eddine,
  • Kamal Bella,
  • Azidine Guezzaz,
  • Said Benkirane,
  • Mourade Azrour,
  • Youssef Kerfi

摘要

The Internet of Things (IoT) has greatly improved connectivity and automation, but its rapid growth has introduced significant security vulnerabilities. To tackle these issues, this study proposes an Intrusion Detection System (IDS) tailored for IoT environments. The proposed model incorporates a Radial Basis Function Neural Network (RBFNN) for precise classification, the Whale Optimization Algorithm (WOA) for feature selection, and an Autoencoder for robust outlier detection. The model was trained and assessed using the NF-ToN-IoT and NF-Bot-IoT datasets, creating a combined dataset (NF-IoT) and an upgraded version (NF-IoT-v2) featuring a two-layer outlier detection strategy. The experimental results show that the proposed IDS achieves 97.25% accuracy (ACC) and 85.77% Matthews Correlation Coefficient (MCC), demonstrating its efficacy in both anomaly detection and classification. These findings underscore the model’s effectiveness in securing IoT networks against cyber threats. Future improvements include enhancing the feature selection method and investigating alternative outlier detection strategies to further optimize model performance.