Scholarlink

<p>Given the abundance of cyber-attacks in wireless communications, machine learning can be essential in identifying intrusions. Ensemble learning has been proven to be a well-known technique for boosting performance and reducing variance. In this paper, we propose a simple ensemble supervised machine learning system, which consists of three classifiers: one-dimensional Convolutional Neural Network, FT-Transformer, and XGBoost, to categorize network traffic as benign or malicious. To find a suitable ensemble method for these classifiers, we implemented three ensemble strategies to integrate these classifiers and assessed their performance. The experimental results demonstrate that our proposed ensemble model achieves strong performance in intrusion detection, as measured by various classification metrics. When evaluated on four benchmark datasets, the model consistently delivers high Accuracy, Precision, Recall, and F1-score. The research showcases the viability of various algorithms, the power of the self-attention model, and the significance of ensemble approaches in strengthening cyber-defense strategies and understanding modern cyber threats.</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

CAT: A simple heterogeneous ensemble learning framework for network intrusion detection

  • 期刊论文

  • 发表日期：

    2025年6月14日

查看全文
  • Zheng Zhang,
  • Amit Das,
  • Guan Huang,
  • Sanjeev Baskiyar

摘要

Given the abundance of cyber-attacks in wireless communications, machine learning can be essential in identifying intrusions. Ensemble learning has been proven to be a well-known technique for boosting performance and reducing variance. In this paper, we propose a simple ensemble supervised machine learning system, which consists of three classifiers: one-dimensional Convolutional Neural Network, FT-Transformer, and XGBoost, to categorize network traffic as benign or malicious. To find a suitable ensemble method for these classifiers, we implemented three ensemble strategies to integrate these classifiers and assessed their performance. The experimental results demonstrate that our proposed ensemble model achieves strong performance in intrusion detection, as measured by various classification metrics. When evaluated on four benchmark datasets, the model consistently delivers high Accuracy, Precision, Recall, and F1-score. The research showcases the viability of various algorithms, the power of the self-attention model, and the significance of ensemble approaches in strengthening cyber-defense strategies and understanding modern cyber threats.