Scholarlink

<p>Intrusion Detection Systems (IDS) are essential for securing IoT networks against increasingly sophisticated cyber threats. This paper provides a comparative performance analysis of various machine learning and deep learning models for IoT intrusion detection. The models, including a Voting Classifier, Autoencoder, Hybrid CNN-LSTM with Ensemble Boosting, and AdaBoost with Decision Trees, were evaluated using the UNSW-NB15 dataset. Both full feature sets and selected features were tested to identify the most effective model configurations. Performance metrics such as accuracy and F1-score were employed, with confusion matrices generated for detailed analysis. Results show that the Soft Voting Ensemble achieves the highest accuracy (0.9517) and F1-score (0.9517), outperforming other models in balancing precision and computational efficiency. The Hybrid CNN-LSTM model also proved effective in detecting complex temporal patterns in network traffic. The study demonstrates the potential of hybrid and ensemble models in enhancing intrusion detection in IoT environments.</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 IoT Intrusion Detection: A Comparative Study of Ensemble-Based Models and Deep Learning Techniques

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Sunil Kumar,
  • Nayani Jindal,
  • Srikruthi Neriyanuri,
  • Parv Tandon,
  • Govind Ram Chhimpa

摘要

Intrusion Detection Systems (IDS) are essential for securing IoT networks against increasingly sophisticated cyber threats. This paper provides a comparative performance analysis of various machine learning and deep learning models for IoT intrusion detection. The models, including a Voting Classifier, Autoencoder, Hybrid CNN-LSTM with Ensemble Boosting, and AdaBoost with Decision Trees, were evaluated using the UNSW-NB15 dataset. Both full feature sets and selected features were tested to identify the most effective model configurations. Performance metrics such as accuracy and F1-score were employed, with confusion matrices generated for detailed analysis. Results show that the Soft Voting Ensemble achieves the highest accuracy (0.9517) and F1-score (0.9517), outperforming other models in balancing precision and computational efficiency. The Hybrid CNN-LSTM model also proved effective in detecting complex temporal patterns in network traffic. The study demonstrates the potential of hybrid and ensemble models in enhancing intrusion detection in IoT environments.