Density Based Mean Shift Clustering with Deep Packet Inspection for Robust Network Traffic Analysis

Abstract

The network traffic analysis is an automatic process to organize the network traffic depending on the parameters. In network traffic analysis, first the traffic data is collected later clustering and classification process are carried out to analyze the network traffic. The continuous estimation and evaluation are the challenging task during traffic classification. But the existing methods failed to reduce the time and to improve the clustering accuracy for network traffic analysis. In order to address these issues, a Density based Mean Shift Clustering and Deep Packet Inspection Classification (DMSC-DPIC) technique is introduced in an effective manner. Initially, the density based mean shift clustering model uses Kernel Density Estimation (KDE) principle to cluster the similar data points by discovering the local maxima modes with higher accuracy. During the density based mean shift clustering, the distribution of data points (i.e., probability density function) is calculated with a set of points without any assumptions on its parameters. Secondly, the deep packet inspection (DPI) classification model is introduced for classifying the network traffic as real time and non-real time traffic using payload of data points with minimal time. DPI classification model classifies the data points into separate classes through examining the related points as part of the session. The experimental evaluation of proposed DMSC-DPIC technique obtains the better performance in terms of clustering accuracy, classification time, and communication overhead compared to the state-of-the-art works.