Dense Neural Network Classification Model for Software Defined Network for Fine Grained Traffic Routing and Flow Analysis

Abstract

Traffic classification in SDN environments acts as the centerpiece, greatly enhancing network management, security, and overall performance. In essence, the classification of traffic in the SDN environment involves classifying network traffic into specific classes according to certain criteria type, source, or destination. Through good classification of network traffic, network operators can optimize resources with unprecedented efficiency, leverage better routing, and grant more priority to the essential data flows that improve QoS. This study utilizes an SDN dataset from Kaggle and evaluates the performance of a state-of-the-art classification model. The paper introduces a further improved Deep Dense Neural Network (DDNN) model, optimized with the Adam optimizer, having a remarkable classification accuracy of 90%. In this paper, the Adam optimizer has been adopted because it allows for an adaptive learning rate that improves convergence and stability during training. Besides, this model showed high scores for other metrics: Precision, Recall, and F1-score all exceeded 90%, reflecting the model's ability to classify reliably and with balance. The low network loss indicates that the model not only gives an accurate result but is also consistent in its performance, with very few misclassifications. Fine-grained traffic classification cannot be underestimated in SDN, given the wide gamut that runs applications over SDN networks. The DDNN model handles diverse types of traffic with a lot of efficiency through extracting deep features from complex datasets, hence increasing the capability of SDN to meet the ever-evolving demands in networks.