Scholarlink

<p>The exponential growth of global data and the increasing utilization of big data technologies have heightened concerns regarding information security, particularly in detecting intrusions in large-scale networked systems. Existing intrusion detection methods often suffer from high false alarm rates, limited capability to detect previously unseen attacks, and inefficiencies in real-time processing, which hinder their practical deployment in dynamic network environments. To address these drawbacks, this paper proposes a scalable hybrid intrusion detection framework that integrates an Extreme Learning Machine (ELM) for optimized feature extraction with a modified Support Vector Machine (SVM) classifier for accurate classification. The approach incorporates three key factors temporal progression, user input, and scalability to process large volumes of data efficiently. The system segments the training data into time intervals, applies user feedback for continuous improvement, and employs scalable storage to reduce computational demands. Evaluation is performed using the UNSW-NB15 dataset, which contains a comprehensive mix of modern attack scenarios and normal traffic patterns. Experimental results demonstrate that the proposed method achieves superior performance compared to existing approaches, delivering higher detection accuracy, reduced false alarm rates, and improved training efficiency, making it well-suited for real-time intrusion detection in high-volume data 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

Using machine learning algorithms with extreme learning to identify instances of malicious activity in data collection

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Yanhua Zhong,
  • Yongqiu Liu

摘要

The exponential growth of global data and the increasing utilization of big data technologies have heightened concerns regarding information security, particularly in detecting intrusions in large-scale networked systems. Existing intrusion detection methods often suffer from high false alarm rates, limited capability to detect previously unseen attacks, and inefficiencies in real-time processing, which hinder their practical deployment in dynamic network environments. To address these drawbacks, this paper proposes a scalable hybrid intrusion detection framework that integrates an Extreme Learning Machine (ELM) for optimized feature extraction with a modified Support Vector Machine (SVM) classifier for accurate classification. The approach incorporates three key factors temporal progression, user input, and scalability to process large volumes of data efficiently. The system segments the training data into time intervals, applies user feedback for continuous improvement, and employs scalable storage to reduce computational demands. Evaluation is performed using the UNSW-NB15 dataset, which contains a comprehensive mix of modern attack scenarios and normal traffic patterns. Experimental results demonstrate that the proposed method achieves superior performance compared to existing approaches, delivering higher detection accuracy, reduced false alarm rates, and improved training efficiency, making it well-suited for real-time intrusion detection in high-volume data environments.