Scholarlink

<p><?tk 4?>Intrusion detection systems play a critical role in protecting modern networks against increasingly sophisticated cyberattacks. This study evaluates the performance of optimized machine learning models for intrusion detection using the CICIDS2017 dataset. Several well-established classifiers are examined for both binary and multi-class attack detection, with appropriate preprocessing and hyperparameter optimization applied to address data imbalance and improve classification reliability. In addition, an ensemble approach combining complementary models is investigated to enhance detection robustness. Experimental results show that tree-based models and ensemble strategies achieve strong overall detection performance across diverse attack categories. These findings support the deployment of data-driven intrusion detection solutions that can assist security operators and infrastructure planners in improving risk monitoring, early threat identification, and informed decision-making in operational network environments.</p> Graphical Abstract <p></p> <p>The graphical abstract visually summarizes the AI-powered Intrusion Detection System proposed in this study. It begins with raw network traffic from the CICIDS2017 dataset, which undergoes preprocessing, including standardization, PCA for dimensionality reduction, and SMOTE for handling class imbalance. The data then feeds into five optimized machine learning classifiers (Logistic Regression, SVM, Decision Tree, Random Forest, and KNN), followed by the integration of a hybrid ensemble model combining KNN and Random Forest to boost accuracy and minimize false positives. The lower segment of the diagram highlights real-time attack classification into categories like DoS, Web Attack, Brute Force, and Benign, enabling proactive cybersecurity responses. This abstract effectively illustrates the system’s end-to-end workflow and its contribution to enhancing real-time threat detection and decision-making accuracy in dynamic network 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

Optimized Machine Learning and Ensemble-Based Intrusion Detection for Binary and Multi-Class Cyber Attack Classification

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • Amer Mahdi F Al Khulayf,
  • Murdhy A. Aldawsari,
  • Atef Ismail,
  • Basma M. Hassan

摘要

Intrusion detection systems play a critical role in protecting modern networks against increasingly sophisticated cyberattacks. This study evaluates the performance of optimized machine learning models for intrusion detection using the CICIDS2017 dataset. Several well-established classifiers are examined for both binary and multi-class attack detection, with appropriate preprocessing and hyperparameter optimization applied to address data imbalance and improve classification reliability. In addition, an ensemble approach combining complementary models is investigated to enhance detection robustness. Experimental results show that tree-based models and ensemble strategies achieve strong overall detection performance across diverse attack categories. These findings support the deployment of data-driven intrusion detection solutions that can assist security operators and infrastructure planners in improving risk monitoring, early threat identification, and informed decision-making in operational network environments.

Graphical Abstract

The graphical abstract visually summarizes the AI-powered Intrusion Detection System proposed in this study. It begins with raw network traffic from the CICIDS2017 dataset, which undergoes preprocessing, including standardization, PCA for dimensionality reduction, and SMOTE for handling class imbalance. The data then feeds into five optimized machine learning classifiers (Logistic Regression, SVM, Decision Tree, Random Forest, and KNN), followed by the integration of a hybrid ensemble model combining KNN and Random Forest to boost accuracy and minimize false positives. The lower segment of the diagram highlights real-time attack classification into categories like DoS, Web Attack, Brute Force, and Benign, enabling proactive cybersecurity responses. This abstract effectively illustrates the system’s end-to-end workflow and its contribution to enhancing real-time threat detection and decision-making accuracy in dynamic network environments.