Scholarlink

<p>Accurate classification of encrypted network traffic is essential for secure and efficient network management, particularly in distributed and heterogeneous Software-Defined Networking (SDN) environments. Traditional traffic classification techniques have become obsolete due to the widespread use of encrypted protocols such as TLS and QUIC. While machine learning methods offer promising alternatives, they often struggle with data scarcity and domain shifts in real-world deployments, especially when a new SDN domain lacks sufficient labeled data. To address this, we propose a novel deep transfer learning framework called DMTE (Deep Multi-source TrAdaBoost Enhanced), which combines deep learning base classifiers with adaptive boosting and an outlier filtering strategy to facilitate cross-domain knowledge transfer. DMTE is integrated within a Knowledge-Defined Networking (KDN) architecture to support collaborative learning across distributed SDN domains. Experimental evaluations on a real-world encrypted traffic dataset demonstrate that DMTE achieves up to 23% improvement in classification accuracy over existing approaches, particularly under limited-data conditions. This work contributes a robust and scalable solution for encrypted traffic classification in intelligent and adaptive networking systems.</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

Multi-source Adaptive Boosting for Encrypted Traffic Classification in Heterogeneous SDNs

  • 期刊论文

  • 发表日期：

    2026年5月28日

查看全文
  • Nam-Thang Hoang,
  • Cong-Son Duong,
  • Tuan Nguyen,
  • Hai-Anh Tran,
  • Truong X. Tran

摘要

Accurate classification of encrypted network traffic is essential for secure and efficient network management, particularly in distributed and heterogeneous Software-Defined Networking (SDN) environments. Traditional traffic classification techniques have become obsolete due to the widespread use of encrypted protocols such as TLS and QUIC. While machine learning methods offer promising alternatives, they often struggle with data scarcity and domain shifts in real-world deployments, especially when a new SDN domain lacks sufficient labeled data. To address this, we propose a novel deep transfer learning framework called DMTE (Deep Multi-source TrAdaBoost Enhanced), which combines deep learning base classifiers with adaptive boosting and an outlier filtering strategy to facilitate cross-domain knowledge transfer. DMTE is integrated within a Knowledge-Defined Networking (KDN) architecture to support collaborative learning across distributed SDN domains. Experimental evaluations on a real-world encrypted traffic dataset demonstrate that DMTE achieves up to 23% improvement in classification accuracy over existing approaches, particularly under limited-data conditions. This work contributes a robust and scalable solution for encrypted traffic classification in intelligent and adaptive networking systems.