Robust ensemble learning technique for traffic classification in SDN networks

By constantly changing flow rules, software-defined network (SDN) offers centralized control over a network of programmable switches. This opens the door for the network to be controlled dynamically and independently. SDN requires information from traffic categorization techniques for the appropriate group of rules to be apply to the proper set of traffic flows. Machine learning nowadays uses a range of categorization methods. A framework known as ensemble that mixes independent models to enhance an overall result has grown in popularity in recent studies showing that applying any algorithm does not always result in the best results for a dataset. Therefore, this paper suggests utilizing the ensemble model with two layers of learning methods to categorize incoming network traffic so that SDN may select the best set of possible traffic regulations using Orange platform. We also apply five machine learning methods and analyze their classification performance in terms of accuracy, precision, and recall. The experimental results reveal that ensemble model-based network traffic classifiers outperform other classifiers based on the proposed framework and the real-world network traffic dataset. Notably, the XGBoost model achieves the best classification performance in every type of traffic examined.