An Intelligent Traffic Classification in SDN-IoT: A Machine Learning Approach

Abstract

In recent years, there has been a sharp increase in IoT devices. Majority of these IoT devices have strict QoS requirements. This has made it very difficult for network providers to provide good network solutions whiles keeping cost in check. To meet the QoS demands in IoT networks, a new paradigm, SDN-IoT, leveraging the advantages of SDN architecture on IoT networks have been proposed to improve network quality. The programmability of the SDN controller allows the application of Machine learning in networks. This paper proposes a Machine learning model that classifies traffic in SDN-IoT networks for traffic engineering. The classification process compares the random forest algorithm, decision tree algorithm, and the K-nearest neighbors’ algorithm. The paper also compares the impact of two feature selection methods, Sequential Feature Selection (SFS) and Shapley additive explanations (SHAP) on the accuracies of the classifiers to reduce the number of features needed for classification. The algorithms are accessed based on their accuracy and F1 score. The best performing algorithm is random forest classifier with SFS which achieves accuracy of 0.833 with six features.