<![CDATA[Classifying network traffic is the foundational step in analyzing diverse applications reliant on network infrastructure, particularly focusing on identifying Over-The-Top (OTT) application traffic using encryption. This methodology empowers Internet service providers and network operators to manage Quality of Service (QoS) performance effectively. Nonetheless, widespread encryption protocols have rendered traditional traffic identification obsolete. Despite limited work in this area, deep learning algorithms are expected to provide a practical solution. This paper introduces a framework outlining the construction of a classifier architecture through the Design Science Research Methodology (DSRM), suitable for producing information system artifacts. The classifier model is built upon deep learning algorithms—CNN, LSTM, and Bi-LSTM. Applying the DSRM approach to deploy the OTT classifier incorporates a deep learning model, offering performance assessment in terms of accuracy, recall, precision, f1-score, and the AUC-ROC curve. The evaluation results of the three models demonstrated strong performance, with accuracy values ranging from 0.83 to 0.96 on the test data. Specifically, the LSTM model show better performance in classifying OTT applications network traffic, achieving an accuracy of 0.96 and an f1-score of 0.95, surpassing the Bi-LSTM and CNN models.]]>
Copyrights © 2024
