EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Machine Learning Approach to Intrusion Detection in Software Defined Networks

Touseeq Danish, Sohan Lal Gupta, Pradhnya Wankhade

International Journal of Information Technology & Computer Sciences Perspectives · 30.06.2026 · 15 (1) · 37-45

DOI: 10.67591/itcs2026.15.1.61 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Information Technology & Computer Sciences Perspectives, online erschienen am 30.06.2026. Verfasst von Touseeq Danish und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objectives

To enhance the detection and mitigation of Distributed Denial of Service (DDoS) attacks especially Low-Rate DDoS (LR-DDoS) attacks, this study aims to propose a machine learning based intrusion detection framework for Software Defined Networks (SDNs). The research goals are to improve accuracy of attack detection, minimize the number of false alarms, speed the response time, and bolster the security and resilience of SDN environments. Theoretical Framework: This study is based on the concepts of Software Defined Networking (SDN), Network Intrusion Detection Systems (NIDS), anomaly-based intrusion detection, entropy-based traffic analysis, and machine learning in cybersecurity. It combines statistical traffic analysis and supervised machine learning to create a multi-layered intrusion detection architecture that is able to detect malicious traffic on the network without compromising normal network operations. Approach: Methodology used in the study is an experimental research approach that is used to design and evaluate a two-tier intrusion detection framework. The first layer consisted of entropy-based traffic analysis to detect suspicious network flows and the second layer used C4.5 decision tree algorithm for classification of malicious and legitimate traffic based on multiple flow-level statistical features. The CIC-DDoS2019 dataset, which consists of normal traffic and various categories of Low-Rate DDoS attacks, was used for model development and evaluation. The proposed framework has been additionally validated in a Software Defined Networking (SDN) environment with Mininet, POX controller with OpenFlow switches, Apache Web Server, Scapy, BoNeSi and Wireshark. Precision, recall, F-score, attack detection time, response time, CPU utilization and flow management efficiency were used to evaluate performance.

Results

The experimental results show that the proposed two tiers detection framework can identify the DDoS attacks at early stage with lower false positive rate and better classification performance. The C4.5 classifier considerably improves the flow level feature analysis of attack classification while entropy-based analysis quickly identifies suspicious traffic. The proposed Traffic Flow Attack Detection (TFAD) mechanism was able to decrease the average attack detection time from about 22 seconds to 12–14 seconds, where the SYN proxy detected attacks in an average time of 6.4 seconds and the ACK proxy detected attacks in 12.6 seconds. Additionally, the integrated framework reduced the error rate, precision, and recall, and achieved high F-score with an efficient controller performance in attack cases. Research Implications: From the practical standpoint, the proposed framework illustrates the potential of integrating statistical analysis with machine learning techniques to enhance real-time intrusion detection and DDoS mitigation in programmable network environments, offering valuable insights for cybersecurity practitioners, SDN administrators, and network engineers. The study also paves the way for future work on intelligent SDN security architectures with advanced machine learning and deep learning approaches. Value: This study contributes to Software Defined Network security by proposing and experimentally validating a hybrid intrusion detection architecture that combines entropy-based anomaly detection with machine learning-based traffic classification. The integration of statistical traffic analysis, flow-based feature engineering, and SDN programmability provides an effective and scalable solution for early DDoS attack detection while improving the accuracy and operational efficiency of next-generation network intrusion detection systems.

Zitieren

APA

Danish, T., Gupta, S. L., & Wankhade, P. (2026). A Machine Learning Approach to Intrusion Detection in Software Defined Networks. International Journal of Information Technology & Computer Sciences Perspectives, 15(1), 37-45. https://doi.org/10.67591/itcs2026.15.1.61

Vancouver

Danish T, Gupta SL, Wankhade P. A Machine Learning Approach to Intrusion Detection in Software Defined Networks. International Journal of Information Technology & Computer Sciences Perspectives. 2026;15(1):37-45. doi:10.67591/itcs2026.15.1.61

BibTeX
@article{Danish2026,
  author = {Danish, Touseeq and Gupta, Sohan Lal and Wankhade, Pradhnya},
  title = {A Machine Learning Approach to Intrusion Detection in Software Defined Networks},
  journal = {International Journal of Information Technology & Computer Sciences Perspectives},
  year = {2026},
  volume = {15},
  number = {1},
  pages = {37--45},
  doi = {10.67591/itcs2026.15.1.61}
}
Kostenlos starten →