Data Traffic Classification in Software Defined Networks (SDN) using supervised-learning

Abstract

Abstract Traffic classification with accuracy is of prime importance in network activities such as security monitoring, traffic engineering, fault detection, accounting of network usage, billing and for providing differentiation in Quality of Service (QoS) parameters of the various network services. Network Traffic Classification is significant in recent days due to rapid growth in the number of internet consumers. The different primitive techniques of network traffic classification have failed to provide reliable accuracy because of 1000 fold scaling in the amount of devices as well as flows. To overcome this drawback, the integration of Software Defined Network (SDN) architecture and machine learning technology is proposed in this paper. Three different supervised learning models, namely Support Vector Machine (SVM), nearest centroid and Naive Bayes (NB), are applied to classify the data traffic based on the applications in a software-defined network platform. The network traffic traces are captured and flows features are generated, which is sent to the classifier for prediction. The accuracy obtained for SVM is 92.3%, NB is 96.79% and the nearest centroid is 91.02%. The challenges faced are in the live network data traffic capture and classification of the applications in the SDN platform.