4G NETWORK IP TRAFFIC CLASSIFICATIONS

Author (s) : K.Karnan and M.Ushanandhini

---

Abstract :
In today's world, the number of Internet services and users is growing rapidly. This leads to a significant increase in internet traffic. The task of classifying IP traffic is so important for ISPs or ISPs as well as various government and private organizations for better network management and security. IP traffic classification involves identifying user activity using network traffic passing through the system. It also helps improve network performance. The use of traditional IP traffic classification mechanisms based on looking at packet loads and port numbers has been drastically reduced because many Internet applications today use dynamic port numbers instead of known port numbers. In addition, there are several encryption techniques today that prevent packet payload inspection: Currently. Various machine learning techniques are commonly used to classify IP traffic. However, the classification of IP traffic in a 4G network has not been studied much. In this research, we developed a new dataset by capturing packets from real-time internet traffic data of a 4G network using a tool called Wire shark. We then extract the inferred properties from the captured packets using Java. We then applied five machine learning models, incl the decision tree. Support for IP traffic classification with Vector Machines, K Nearest Neighbors, Random Forest and Naive Bayes. Random Forest was found to give the best accuracy of about 87%.

Keywords : Vector Machines, Decision tree

Conference Date : 30/08/2019

Pages : 33

Paper ID : 19031