NETWORK TRAFFIC ANALYSIS USING MACHINE LEARNING

Abstract

In the field of networking, it is often necessary to identify the sorts of applications that traverse a
network in order to complete certain tasks. Classification of network traffic is primarily used by Internet
service providers (ISPs) to examine the characteristics necessary for network architecture, which
impacts the network's overall performance. There are several classification methods for network
protocols, such as port-based, payload-based, and Machine Learning-based, and each has its own
advantages and disadvantages. Due to its widespread use in different domains and academics' rising
awareness of its superior accuracy when compared to others, the Machine Learning approach is now
prominent. In this research, we examine the performance of two fundamental algorithms, Nave Bayes
and K nearest, when applied to networking data retrieved from live video stream using the Wireshark
program. Python's sklearn library is used to implement the Machine Learning algorithm, together with
the numpy and pandas libraries as assistance libraries. Finally, we see that the K closest approach
provides more precise predictions than the Naive Bayes, Decision Tree, and Support Vector Machine
algorithms.