Scholarlink

<p>The rapid development of the Internet of Things (IoT) has enabled efficient information exchange and intelligent resource sharing. However, the widespread deployment of devices and inadequate security measures have significantly increased the risk of network attacks, making intrusion detection systems (IDS) a key research focus. Existing IoT-based IDS heavily rely on hyperparameter optimization, which demands high computational resources and poses challenges for practical deployment. Moreover, IoT data often contain redundancy and high-dimensional features, further increasing complexity and degrading model performance. To address these issues, this study proposes a novel stacking ensemble method, HF-IDS. The method enhances the grey wolf optimization algorithm (GWO) by incorporating nonlinear decreasing control parameters, an elite strategy, and a memory mechanism(EEM-GWO), using the improved EEM-GWO to optimize the hyperparameters of machine learning and deep learning models. Prediction results from individual models are integrated, with final classification performed by random forest. Additionally, a multi-level feature selection technique is introduced, using stepwise optimization based on standard deviation, mutual information, and Pearson correlation coefficient to reduce dimensionality and identify key features. Experiments on the UNSW-NB15, CICIDS2017, and ToN IoT datasets cover both binary and multi-class classification tasks. Results show that HF-IDS achieves accuracy rates of 98.25%, 99.97%, and 99.82% for binary classification, respectively. Compared to the original feature set, the selected set reduces time and resource usage by 20% to 50%. Additionally, the results clearly demonstrate the significant potential of hyperparameter optimization in improving the accuracy and efficiency of IoT-based intrusion detection systems.</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

An ensemble framework with improved grey wolf optimization algorithm and multi-level feature selection for IoT intrusion detection

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Kexin Wu,
  • Yueqin Li,
  • Jixu Sun,
  • Qiurong Qin,
  • Jinlong Li

摘要

The rapid development of the Internet of Things (IoT) has enabled efficient information exchange and intelligent resource sharing. However, the widespread deployment of devices and inadequate security measures have significantly increased the risk of network attacks, making intrusion detection systems (IDS) a key research focus. Existing IoT-based IDS heavily rely on hyperparameter optimization, which demands high computational resources and poses challenges for practical deployment. Moreover, IoT data often contain redundancy and high-dimensional features, further increasing complexity and degrading model performance. To address these issues, this study proposes a novel stacking ensemble method, HF-IDS. The method enhances the grey wolf optimization algorithm (GWO) by incorporating nonlinear decreasing control parameters, an elite strategy, and a memory mechanism(EEM-GWO), using the improved EEM-GWO to optimize the hyperparameters of machine learning and deep learning models. Prediction results from individual models are integrated, with final classification performed by random forest. Additionally, a multi-level feature selection technique is introduced, using stepwise optimization based on standard deviation, mutual information, and Pearson correlation coefficient to reduce dimensionality and identify key features. Experiments on the UNSW-NB15, CICIDS2017, and ToN IoT datasets cover both binary and multi-class classification tasks. Results show that HF-IDS achieves accuracy rates of 98.25%, 99.97%, and 99.82% for binary classification, respectively. Compared to the original feature set, the selected set reduces time and resource usage by 20% to 50%. Additionally, the results clearly demonstrate the significant potential of hyperparameter optimization in improving the accuracy and efficiency of IoT-based intrusion detection systems.