Computer Network Analysis by Visualization

Abstract

The explosive growth of Internet has raised interest in traffic analysis. Understanding what traffic traverse the network is important for operation, investments, research and design of new protocols. However network traffic analysis has not evolved as rapidly as network usage. Many researchers still look at data in raw text format even though the human brain is much better at pattern recognition in images than text. The purpose with this thesis is to develop a tool that builds a graph to visualize network traffic. Network analysis using this approach is not new, but there are no tools available where visualization is the focus. The network graph is built by defining hosts as nodes and communication between hosts as edges. To enable analysis the user can select a subset of the traffic to visualize. The tool is able to produce graphs on large data sets, 3000 nodes and 30000 edges, on a home computer. We have tested the tool on generated data and on data provided by the MonNet project. The tool is ready for testing but further development is needed since the graph library we used is resource intensive when visualizing large graphs.