Scholarlink

This research investigates the application of TabNet, a deep learning model optimized for tabular data, for binary classification of Software Defined Networking (SDN) traffic. Using a combination of normal and Open vSwitch (OVS) traffic datasets, the model was trained to differentiate between benign and potentially anomalous traffic patterns. The preprocessing pipeline included feature selection, normalization, and encoding, followed by model training using PyTorch TabNet. The classifier achieved a high validation accuracy of 99.51%, with strong precision and recall, highlighting its capability for real-time anomaly detection in SDN environments. Visualizations such as ROC curves, precision-recall plots, confusion matrices, and PCA projections further validated the model’s effectiveness and interpretability.

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 Expository Deep Learning Approach for Network Intrusion Detection in Software Defined Networking

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • E. C. Nwokorie,
  • D. O. Njoku,
  • M. E. Nwanga,
  • S. A. Okolie,
  • C. D. Anyiam,
  • J. E. Jibiri,
  • C. G. Onukwugha,
  • I. H. Ajunwa,
  • C. O. Amadi,
  • F. O. Nwokoma,
  • A. I. Otuonye,
  • U. C. Onyemauche

摘要

This research investigates the application of TabNet, a deep learning model optimized for tabular data, for binary classification of Software Defined Networking (SDN) traffic. Using a combination of normal and Open vSwitch (OVS) traffic datasets, the model was trained to differentiate between benign and potentially anomalous traffic patterns. The preprocessing pipeline included feature selection, normalization, and encoding, followed by model training using PyTorch TabNet. The classifier achieved a high validation accuracy of 99.51%, with strong precision and recall, highlighting its capability for real-time anomaly detection in SDN environments. Visualizations such as ROC curves, precision-recall plots, confusion matrices, and PCA projections further validated the model’s effectiveness and interpretability.