Precise traffic classification is essential for network operations like traffic engineering, fault detection, network accounting, invoicing, and differentiating the quality of service (QoS) characteristics of distinct network services. The importance of network traffic classification has increased recently as a result of the sharp rise in internet users. The various archaic methods of classifying network traffic have not been able to deliver consistent accuracy due to the increase in both the number of devices and flows. This study proposes the merging of machine learning (ML) technology and software defined network (SDN) architecture to solve this issue. In a software-defined network platform, three distinct supervised learning models support vector machine (SVM), nearest centroid (NC), and Naïve Bayes (NB) are utilized to categorize data flow according to its uses. After capturing the network traffic traces, flows features are created, and supplied to the classifier for prediction. SVM achieved 93.2% accuracy, NB 97.69% and the NC 92.01%. The SDN platform's live network data traffic capture and application classification present issues.
Keywords
Naïve Bayes; Nearest centroid; Prediction; Software defined network; Support vector machine
