Abstract
Accurate traffic categorization is needed for security monitoring, traffic engineering, fault diagnosis, accounting for network consumption, pricing, and distinguishing the Quality of Service (QoS) of various network services. Classifying network traffic has become a critical problem due to the growth of internet users. With a thousand-fold increase in devices and flows, rudimentary techniques to network traffic classification have failed. In this study, researchers propose a method that combines SDN architecture with machine learning tools to solve the problem. This action is taken to counteract the drawback at hand. There are three main supervised learning models used in a software-defined network architecture to categorise traffic levels of the data dependent on the apps it serves. To name a few, we have models such as the Naive Bayes (NB) classifier, the closest proposed centroid method, and the Support Vector Machine (SVM). After the network traffic traces have been gathered and the flows characteristics have been formed, they are delivered to the classifier so that a prediction may be made. The accuracy that was acquired using SVM was found to be 98.21%, whereas the accuracy gained using NB was found to be 94.29% and proposed centroid method was found to be 99.98%. The difficulties lie in the real network data traffic capturing as well as the categorization of the applications inside the SDN platform.
Keyword
Software Define Network, Naïve Bayes , Support Vector Machine, Quality of Service, Traffic Classification.
PDF Download (click here)
|