文章基本信息

标题：NETWORK TRAFFIC CLASSIFICATION USING ENSEMBLE LEARNING WITH TIME RELATED FEATURES
本地全文：下载
作者：Muhammad Anwarul Azim ; Tanvir Mohammad Khairul Islam
期刊名称：International Journal of New Computer Architectures and their Applications
印刷版ISSN：2220-9085
出版年度：2020
卷号：10
期号：2
页码：23-31
DOI：10.17781/P002674
语种：English
出版社：Society of Digital Information and Wireless Communications
摘要：Network traffic classification has become important with the rapid growth of the Internet and online applications. Though, there were researches that applied different machine learning algorithms for traffic classification purposes, the continuous expansion of technologies and applications in stationary and mobile are creating a dynamic environment. Because of encryption in today’s Internet, traffic classification still poses a great deal of concern for researchers and network communities. This work proposes ensemble learning including Voting, Bagging, and Boosting for traffic classification, and then compares them with their own base classifiers when used individually. Time-related features are focused which are independent of data encryption on the UNB ISCX dataset, containing flow duration, inter-arrival time, byte rate, packet rate, etc. Among different techniques, Random Forest outperforms nearly all others with respect to various evaluation matrices such as accuracy, precision, recall, and f1-score. In the case of VPN traffic and Non-VPN traffic, it gives almost 90.65% and 95.42% accuracy respectively. In the case of combined VPN and Non-VPN traffic, we achieve 90.18% accuracy for classifying traffic categories which is a significant improvement from previous works.