Network Traffic Classification Using Supervised Learning Algorithms

Abstract

Network traffic classification is crucial for traffic monitoring and application-based policy enforcement. However, the widespread use of encrypted protocols has greatly challenged conventional traffic classification techniques using packet payload and port numbers. For the network application in this paper, two machine learning algorithms, Decision Tree (DT) and Random Forest (RF) are used. An open-access Kaggle dataset with six different types of applications is used for this study. To achieve the best values for model training, loop iteration is used rather than the hyper-parameter optimization technique. When compared to DT, RF has the highest accuracy (99.72%). In order to improve the classification process and various hidden patterns connected with the statistical features, more statistical features were taken into account in comparison to other related works that had already been done. The outcomes demonstrate the potency of supervised learning algorithms for categorizing network traffic.