Scholarlink

In Software Defined Networks (SDNs), the controller is a software application that manages flow control for improved network management. A massive amount of information is generated at the controller which could pose many security threats and potential targets to SDN. Nowadays, the severity of Distributed Denial of Services (DDoS) attacks grows enormously. Machine learning (ML) approaches have the ability to progressively learn the traffic pattern from the generated data at the controller. In spite of many works, accurate classifier selection for the attack detection is an open question. The major contribution of this research is the use K-nearest neighbor (KNN) classifier with KD-tree data structure (KD-KNN). The KD-tree approach tries to improve the detection accuracy as well as running time by decreasing the amount of time while calculating the Euclidean distance between two data points. Experimental results with proper analysis have been conducted over an OpenFlow traced DDoS data set. The experimental results on the customized dataset show that at \(k=5\) , KD-KNN achieves 98.96% accuracy and the average execution time is 13.84 s as compared to 16.51 s for KNN.

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

Toward Detecting DDoS Attacks in SDN with KD-KNN

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Debashis Sahoo,
  • Kshira Sagar Sahoo,
  • Rakesh Chandra Balabantaray,
  • Arati Behera,
  • Anamay Sarkar,
  • Monowar Bhuyan

摘要

In Software Defined Networks (SDNs), the controller is a software application that manages flow control for improved network management. A massive amount of information is generated at the controller which could pose many security threats and potential targets to SDN. Nowadays, the severity of Distributed Denial of Services (DDoS) attacks grows enormously. Machine learning (ML) approaches have the ability to progressively learn the traffic pattern from the generated data at the controller. In spite of many works, accurate classifier selection for the attack detection is an open question. The major contribution of this research is the use K-nearest neighbor (KNN) classifier with KD-tree data structure (KD-KNN). The KD-tree approach tries to improve the detection accuracy as well as running time by decreasing the amount of time while calculating the Euclidean distance between two data points. Experimental results with proper analysis have been conducted over an OpenFlow traced DDoS data set. The experimental results on the customized dataset show that at \(k=5\) , KD-KNN achieves 98.96% accuracy and the average execution time is 13.84 s as compared to 16.51 s for KNN.