Scholarlink

<p>Securing Internet of Things (IoT) systems has become increasingly challenging due to the rapid evolution of cyber-attacks. Intrusion Detection Systems (IDS) play a vital role in safeguarding networks against diverse threats and vulnerabilities; however, challenges related to class imbalance and computational efficiency persist. In this study, we propose a hybrid IDS framework that integrates the Harmony Search Algorithm (HSA) for optimized feature selection with Adaptive Synthetic Sampling (ADASYN) to address class imbalance and a machine learning (ML) model to classify genuine traffic from attacks. We evaluated the framework using Bagging, Gradient Boosting (GB), LightGBM (LGBM), and Random Forest (RF) classifiers. The experimental results indicate that the RF-HSA-ADASYN approach is more accurate than baseline methods, achieving 96<i>.</i>92 ± 0<i>.</i>18% accuracy on UNSW-NB15 and 99<i>.</i>94 ± 0<i>.</i>03% accuracy on CIC-IoT23, significantly outperforming the state-of-the-art models. The proposed framework systematically integrates statistical validation using ANOVA in conjunction with inference time analysis to ensure both analytical rigor and computational efficiency. The framework offers a promising solution for real-time threat detection in dynamic network environments while maintaining adaptability to evolving attack vectors.</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

A hybrid HSA-ADASYN framework for robust intrusion detection in imbalanced IoT environments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月22日

查看全文
  • Tibebu Bekele Shana,
  • Mayank Agarwal,
  • Ebenezer Esenogho,
  • Samrat Mondal

摘要

Securing Internet of Things (IoT) systems has become increasingly challenging due to the rapid evolution of cyber-attacks. Intrusion Detection Systems (IDS) play a vital role in safeguarding networks against diverse threats and vulnerabilities; however, challenges related to class imbalance and computational efficiency persist. In this study, we propose a hybrid IDS framework that integrates the Harmony Search Algorithm (HSA) for optimized feature selection with Adaptive Synthetic Sampling (ADASYN) to address class imbalance and a machine learning (ML) model to classify genuine traffic from attacks. We evaluated the framework using Bagging, Gradient Boosting (GB), LightGBM (LGBM), and Random Forest (RF) classifiers. The experimental results indicate that the RF-HSA-ADASYN approach is more accurate than baseline methods, achieving 96.92 ± 0.18% accuracy on UNSW-NB15 and 99.94 ± 0.03% accuracy on CIC-IoT23, significantly outperforming the state-of-the-art models. The proposed framework systematically integrates statistical validation using ANOVA in conjunction with inference time analysis to ensure both analytical rigor and computational efficiency. The framework offers a promising solution for real-time threat detection in dynamic network environments while maintaining adaptability to evolving attack vectors.