Software defined networking based network traffic classification using machine learning techniques

Area of Science:

• Computer Science
• Network Engineering
• Machine Learning

Background:

• Conventional network traffic classification methods are failing due to encrypted and dynamic traffic.
• Existing Software Defined Networking (SDN) and Machine Learning (ML) approaches have limitations in accuracy and real-time detection.
• There is a need for robust methods to classify diverse network traffic types effectively.

Purpose of the Study:

• To evaluate the effectiveness of various supervised and unsupervised Machine Learning (ML) models for network traffic classification within an SDN environment.
• To compare the performance of different ML algorithms in classifying Domain Name System (DNS), Telnet, Ping, and Voice traffic.
• To demonstrate an integrated ML-SDN approach for accurate and efficient real-time traffic classification.

Main Methods:

• Simulated network traffic using the Distributed Internet Traffic Generator (D-ITG) tool.
• Implemented supervised (Logistic Regression, Decision Tree, Random Forest, AdaBoost, Support Vector Machine) and unsupervised (K-means clustering) ML models.
• Utilized Software Defined Networking (SDN) implemented in Mininet for network architecture and traffic generation, with classification performed in an Anaconda Python environment.

Main Results:

• The Decision Tree supervised learning model achieved the highest classification accuracy at 99.81%.
• The proposed ML-SDN integration demonstrated superior performance compared to other tested algorithms for both offline and real-time traffic.
• The approach effectively classified various traffic types, including encrypted packets.

Conclusions:

• Integrating Machine Learning (ML) with Software Defined Networking (SDN) offers an efficient and accurate solution for network traffic classification.
• This approach enhances Quality of Service (QoS), enables detection of encrypted packets, and supports deep packet inspection.
• The Decision Tree model shows significant promise for real-time network traffic management and security.