Evaluating Machine Learning Algorithms for Effective Network Protocol Classification

Abstract

The current study illustrates the effectiveness of machine learning for the classification of protocols. Many critical operations on the network need to be observed, such as traffic analysis, quality of services, and traffic optimization. Given the emerging complexity of the network environment, it has become a challenge for a traditional classifier to deal with encrypted traffic and dynamic port assignment by data traffic. In the current study, three machine learning models were used and examined, named Decision Tree (TD), Random Forest (RF), and Naive Bayes(NB), which were evaluated based on metrics such as precision, precision, recall, and F1 score. The results indicated that both the Random Forest and the Decision tree outperform the NB, the highest achievement of the accuracy was for Random Forest with 96 %. This work shows the potential of using machine learning for the management of modern networks and provides the foundation for further studies