Fitting a code-red virus spread model: An account of putting theory into practice

This paper is about fitting a model for the spreading of a computer virus to measured data, contributing not only the fitted model, but equally important, an account of the process of getting there. Over the last years, there has been an increased interest in epidemic models to study the speed of virus spread. But parameterising such models is hard, because due to the unexpected nature of real outbreaks, there is not much solid measurement data available, and the data may often have imperfections. We propose a mean-field model for computer virus spread, and use parameter fitting techniques to set the model's parameter values based on measured data. We discuss a number of steps that had to be taken to make the fitting work, including preprocessing and interpreting the measurement data, and restructuring the model based on the available data. We show that the resulting parameterised model closely mimics real system behaviour, with a relative squared error of 0.7%.