Scholarlink

This research’s aim is to address the critical issue of class imbalance in intrusion detection systems (IDS) in software-defined networking environments. This paper introduces a novel approach that exploits advanced deep learning techniques to improve minority class attack detection, often missed because they are rare. Balancing the dataset using data synthesis with GAN and SMOTE, this study allows different classifiers to improve their performance. The research explores the effectiveness of multiple deep learning architectures, including MLPs, CNNs, and SNNs, in detecting intrusions. The results show that GAN-based augmentation significantly outperforms traditional methods such as SMOTE, reducing false negatives and increasing overall detection accuracy.The paper also places an emphasis on the preprocessing technique of data that will include mean imputation as well as standardization techniques to enhance the input quality. Results show how the proposed integrated approach is able to improve not only the accuracy of intrusion detection but also the whole security framework in SDN environments.

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

Optimizing Class Imbalance and Enhancing Intrusion Detection in SDN Environments Using Deep Learning Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • K. Suresh Babu,
  • M. Sireesha,
  • T. G. Ramnadh Babu,
  • G. Venkatesh,
  • D. Sreenivas,
  • M. Venkatesh

摘要

This research’s aim is to address the critical issue of class imbalance in intrusion detection systems (IDS) in software-defined networking environments. This paper introduces a novel approach that exploits advanced deep learning techniques to improve minority class attack detection, often missed because they are rare. Balancing the dataset using data synthesis with GAN and SMOTE, this study allows different classifiers to improve their performance. The research explores the effectiveness of multiple deep learning architectures, including MLPs, CNNs, and SNNs, in detecting intrusions. The results show that GAN-based augmentation significantly outperforms traditional methods such as SMOTE, reducing false negatives and increasing overall detection accuracy.The paper also places an emphasis on the preprocessing technique of data that will include mean imputation as well as standardization techniques to enhance the input quality. Results show how the proposed integrated approach is able to improve not only the accuracy of intrusion detection but also the whole security framework in SDN environments.