Scholarlink

<p>In the digital age, cybersecurity plays a critical role as organizations and individuals face serious threats from cyber attacks. Consequently, the importance of Intrusion Detection Systems (IDS) is continually growing. Feature Selection (FS) plays a critical role in enhancing the performance of IDS. We introduce a filter-based feature selection method that leverages Mutual Information (MI) to assess feature relevance while treating the number of selected features as a distinct objective. To solve this multi-objective optimization problem, we enhance the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) by integrating an MI-based function for initial population generation. Additionally, to achieve superior performance, we implement a modified Response Surface Methodology (RSM) for parameter tuning. To demonstrate the effectiveness of our method, we employ Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM) classifiers on three datasets, NSL-KDD, UKM-IDS20, and UNR-IDD 2023 for both binary and multi-class classification tasks. The results show an accuracy of 99.88% with 29 features using the RF classifier, 99.99% with 13 features using the RF classifier, and 98.62% with 2 features using the RF classifier for the NSL-KDD, UKM-IDS20, and UNR-IDD 2023 datasets respectively for multi-class classification. Comparative analysis with existing research confirms the efficiency of the proposed method.</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 entropy-based multi-objective feature selection method for network intrusion detection

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Zeinab Raeisi,
  • Hamid Reza Maleki,
  • Reza Akbari

摘要

In the digital age, cybersecurity plays a critical role as organizations and individuals face serious threats from cyber attacks. Consequently, the importance of Intrusion Detection Systems (IDS) is continually growing. Feature Selection (FS) plays a critical role in enhancing the performance of IDS. We introduce a filter-based feature selection method that leverages Mutual Information (MI) to assess feature relevance while treating the number of selected features as a distinct objective. To solve this multi-objective optimization problem, we enhance the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) by integrating an MI-based function for initial population generation. Additionally, to achieve superior performance, we implement a modified Response Surface Methodology (RSM) for parameter tuning. To demonstrate the effectiveness of our method, we employ Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM) classifiers on three datasets, NSL-KDD, UKM-IDS20, and UNR-IDD 2023 for both binary and multi-class classification tasks. The results show an accuracy of 99.88% with 29 features using the RF classifier, 99.99% with 13 features using the RF classifier, and 98.62% with 2 features using the RF classifier for the NSL-KDD, UKM-IDS20, and UNR-IDD 2023 datasets respectively for multi-class classification. Comparative analysis with existing research confirms the efficiency of the proposed method.