Classificação de tráfego entrante em uma topologia SDN

Abstract

Computer networks have become a vital tool for transporting information. The use of Software Defined Networks can enable the development of techniques to improve network performance with respect to security, quality of service and traffic engineering. The implementation of these techniques can be facilitated by classifying incoming traffic on the network. This work proposed a comparative study of machine learning algorithms for the classification of incoming traffic in an SDN topology. The classifiers' performance was assessed using the metrics accuracy, precision, recall and f1-score, in addition to the training and validation times of the models. The Random Forest algorithm was considered the most efficient in the considered traffic classification scenario. He achieved values ​​similar to the best results regarding the metrics accuracy, precision, recall and f1-score, but obtained lower values ​​in the training and validation times.