Comparative Analysis and Classification of HTTPS Traffic Using Packet Burst Statistics: An Evaluation of State-of-the-Art Methods

This study examines the classification of HTTPS traffic using packet burst statistics, a crucial aspect of modern internet usage with significant implications for network security, traffic management, and service quality. Leveraging extensive datasets from real backbone networks, we categorize HTTPS traffic into five primary types: Live Video Streaming, Video Player, Music Player, File Uploading/Downloading, and Website & Other Traffic. We employ various machine learning algorithms with particular emphasis on Random Forest and XGBoost, which demonstrate high accuracy rates. Furthermore, we incorporate recent advancements such as the Kolmogorov-Arnold Network (KAN) method for comparative analysis, enhancing the robustness of our study. We present a comprehensive methodology for model performance comparison and clustering analysis. Our findings have practical applications in network security, traffic management, and service quality enhancement. This research contributes significantly to the field, paving the way for future studies aimed at more effective classification and management of HTTPS traffic.