In-Depth Network Traffic Analysis using Machine Learning Perspective: Characterization and Classification

This study provides an in-depth exploration of Network Traffic Analysis (NTA) utilizing a Machine Learning (ML) perspective, focusing on both characterization and classification. The study initiates with a comprehensive examination of traffic behavior, allowing for the identification of patterns and the establishment of correlations among various attributes. Notably, flow duration is identified as a key label demonstrating a positive correlation with cumulative inter-arrival time in the forward direction. Following this, the traffic is classified into distinct categories, facilitating an evaluation of their priority in commu-nication. The comparative study of nine algorithms reveals that while random forest attains superior accuracy, the decision tree offers expedited computational efficiency. This research advances our understanding of NTA methodologies, providing insights into their applicability and effectiveness.