Scholarlink

Software-Defined Networking (SDN) is an emerging paradigm that offers the flexibility to program and manage networks from a centralized point. Its dynamic nature, however, poses several security challenges. In an effort to resolve these issues and enhance traffic isolation, this research proposes a two-stage micro-segmentation technique. In the first stage, a modified Adaptive Resonance Theory 2A (ART2A) clustering algorithm, ART2A-RNorm, is utilized to classify the network traffic flows according to the destination port numbers, forming initial segmentation zones. In the second stage, micro-segmentation is done more precisely with Classless Inter-Domain Routing (CIDR) based on the source IP addresses of each cluster. The modified ART2A demonstrated clear advantages in experimental performance on the UNSW-NB15 dataset, including an increase in Silhouette Score (0.6295 to 0.7718), a decrease in Davies-Bouldin Index (0.7966 to 0.5971), and an improvement in Calinski-Harabasz Index of more than two times. Moreover, CIDR-based microsegmentation produced silhouette values of 0.95–1.0, which nearly fall within the optimal range, developing highly coherent subnet structures. These results highlighted the effectiveness of the method that combines port-based clustering with CIDR-based subdivision for more flexible and stronger segmentation in SDN contexts.

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

A Two-Stage Micro-segmentation Framework for SDN Security Using Modified ART2A Clustering and CIDR

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Manar H. Bashaa,
  • Wesam S. Bhaya,
  • Nabeel H. Kaghed Al-aaraji

摘要

Software-Defined Networking (SDN) is an emerging paradigm that offers the flexibility to program and manage networks from a centralized point. Its dynamic nature, however, poses several security challenges. In an effort to resolve these issues and enhance traffic isolation, this research proposes a two-stage micro-segmentation technique. In the first stage, a modified Adaptive Resonance Theory 2A (ART2A) clustering algorithm, ART2A-RNorm, is utilized to classify the network traffic flows according to the destination port numbers, forming initial segmentation zones. In the second stage, micro-segmentation is done more precisely with Classless Inter-Domain Routing (CIDR) based on the source IP addresses of each cluster. The modified ART2A demonstrated clear advantages in experimental performance on the UNSW-NB15 dataset, including an increase in Silhouette Score (0.6295 to 0.7718), a decrease in Davies-Bouldin Index (0.7966 to 0.5971), and an improvement in Calinski-Harabasz Index of more than two times. Moreover, CIDR-based microsegmentation produced silhouette values of 0.95–1.0, which nearly fall within the optimal range, developing highly coherent subnet structures. These results highlighted the effectiveness of the method that combines port-based clustering with CIDR-based subdivision for more flexible and stronger segmentation in SDN contexts.