Scholarlink

<p>This research presents a novel intrusion detection system (IDS) named as MOGWO-FWEL-SDN—designed to enhance the performance and adaptability of Software Defined Networks (SDNs) in the face of evolving cyber threats, especially within IoT and IIoT environments. The proposed system integrates Multiobjective Grey Wolf Optimization (MOGWO) with Feature Weighting and Ensemble Learning (FWEL) to construct a robust and scalable IDS pipeline. Utilizing four benchmark datasets—KDD99, CIC-IoT 2022, CIC-IoT 2023, and Edge-IIoTset—the model demonstrates its capability across traditional, modern, and industrial IoT attack vectors. Each dataset undergoes unified preprocessing, augmentation, and schema alignment to ensure feature consistency, followed by feature selection using MOGWO based on entropy and Gini index minimization. Selected features are weighted and passed through dual ensemble classifiers using bagging and boosting with Random Forest (RF). Experimental evaluations yield exceptionally high detection performance across datasets, achieving up to 99.4% accuracy on CIC-IoT 2023 and 99.3% on KDD99 and Edge-IIoT datasets respectively. Class-wise analysis reveals F1-scores consistently above 98.2% even for complex attack types like Infiltration, Ransomware, and Privilege Escalation. Ablation studies validate the additive benefits of each module, with feature weighting and optimization contributing up to 1.5% performance gain.</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

Adaptive feature weighted ensemble learning (FWEL) framework for programmable network threat detection using multi-objective grey wolf optimization (MOGWO)

  • 期刊论文

  • 发表日期：

    2025年12月3日

查看全文
  • Rupali Dhir,
  • Kamal Malik

摘要

This research presents a novel intrusion detection system (IDS) named as MOGWO-FWEL-SDN—designed to enhance the performance and adaptability of Software Defined Networks (SDNs) in the face of evolving cyber threats, especially within IoT and IIoT environments. The proposed system integrates Multiobjective Grey Wolf Optimization (MOGWO) with Feature Weighting and Ensemble Learning (FWEL) to construct a robust and scalable IDS pipeline. Utilizing four benchmark datasets—KDD99, CIC-IoT 2022, CIC-IoT 2023, and Edge-IIoTset—the model demonstrates its capability across traditional, modern, and industrial IoT attack vectors. Each dataset undergoes unified preprocessing, augmentation, and schema alignment to ensure feature consistency, followed by feature selection using MOGWO based on entropy and Gini index minimization. Selected features are weighted and passed through dual ensemble classifiers using bagging and boosting with Random Forest (RF). Experimental evaluations yield exceptionally high detection performance across datasets, achieving up to 99.4% accuracy on CIC-IoT 2023 and 99.3% on KDD99 and Edge-IIoT datasets respectively. Class-wise analysis reveals F1-scores consistently above 98.2% even for complex attack types like Infiltration, Ransomware, and Privilege Escalation. Ablation studies validate the additive benefits of each module, with feature weighting and optimization contributing up to 1.5% performance gain.