Virtual user simulation

Abstract

This project is about human–computer interaction, consisting of two parts. The first part is to analyze human behavior through the observation of machine/system behavior – solely through network traffic. By observing this in a multi user system – what they do and when they do it, this thesis aims to see if distinct features of the population can be found and later reenacted. This will be achieved through observing network traffic, gathering it in a structured way and characterizing it. Humans and their computer workstations are the principal factors in making the traffic that is going to be measured. The second part is to simulate the behavior and network traffic observed with virtual users – programs which behave similarly to their real counterparts. This is done with scripting, and results in a configurable tool which can simulate human behavior as seen from the network. Humans are thus eliminated in favor of computer simulated virtual users. The virtual users will try to generate network traffic with the same type, amount, and characteristics as observed in the first part, in a simulation with a hitherto unprecedented level of realism. Ideally, if this is repeated, it will form an idempotent loop ending up still looking like the original system with real users.