Application-Based Traffic Classification by Employing Machine Learning Models on Software-Defined Networking

The exponential growth of new applications and services on the Internet demands increased bandwidth allocation and quality of service. However, traditional networks need help in addressing this issue. Software-defined networking has emerged as a promising solution because it provides a global view of network traffic by separating the control plane from the data plane. This separation offers the necessary flexibility to incorporate new functionalities, including the integration of machine learning, which has become essential for optimizing and effectively managing networks. Therefore, this article aims to define a model for application classification using machine learning techniques to enhance bandwidth allocation and quality of service in Software-Defined Networks. We employ the Sample, Explore, Modify, Model, and Assess (SEMMA) methodology to develop the classifier and train four machine learning models, Decision Trees (DT), Random Forests (RF), Support Vector Machines (SVM), and K-nearest neighbor (KNN), using multiple feature sets from a public data set. The results obtained demonstrate that with 15 features, the highest level of accuracy is achieved, with the following model accuracies: RF (99.99%), DT (99.89%), KNN (99.92%), and SVM (92.06%). These results underscore the remarkable effectiveness of the RF model in application classification.