Deep Learning Approaches for Network Traffic Classification In Software Defined Network

Abstract

The exponential growth of network traffic demands advanced management solutions. Software Defined Networking (SDN) addresses this need by providing a centralized control mechanism to measure, manage, and predict traffic. However, the volume of data processed by the SDN controller is substantial. Machine learning (ML) techniques have emerged as a viable method for analyzing this data. Accurate network traffic classification is vital for effective resource allocation, security, and granular network control. Traditional methods, including port-based analysis and deep packet inspection (DPI), have become less effective due to the computational demands of modern internet applications. This paper proposes a novel deep learning model for SDN environments designed to accurately classify a diverse range of traffic applications with high efficiency. The model demonstrated superior performance compared to conventional methods, achieving an accuracy of 92%, a precision of 92%, a recall of 88%, and an F1-score reflecting this high reliability. The findings suggest promising directions for future research to further enhance traffic classification in software-defined networks.