Network Traffic Classification Using Machine Learning Algorithms

Abstract

Nowadays, Network Traffic Classification has got pivotal significance owing to high growth in the number of internet users. People use a variety of applications while browsing the pages of internet. It is very crucial for internet service providers (ISPs) to keep an eye on the network traffic. Most of the researches made on Network Traffic Classification, using Machine Learning Based Traffic Identification to collect data set from one campus network, don’t provide far better results. In this paper, we attempt to achieve highly precise results using different kinds of data sets and Machine Learning (ML) algorithms. We use two data sets, HIT and NIMS data sets for this work. Firstly, we capture online internet traffic of seven different kinds of applications such as DNS, FTP, TELNET, P2P, WWW, IM and MAIL to make data sets. Then, we extract the features of captured packets using NetMate tool. Thereafter, we apply three ML algorithms Artificial Neural Network, C4.5 Decision Tree and Support Vector Machine to compare the results of each algorithm. Experimental results show that all the algorithms give highly accurate results. But C4.5 decision tree algorithm provides 97.57% highly precise results when compared to other two machine learning algorithms.