%0 Journal Article %A Barath Kumar and Oliver Niggemann and Juergen Jasperneite %D 2010 %J International Journal of Computer and Information Engineering %B World Academy of Science, Engineering and Technology %I Open Science Index 37, 2010 %T Statistical Models of Network Traffic %U https://publications.waset.org/pdf/1610 %V 37 %X Model-based approaches have been applied successfully to a wide range of tasks such as specification, simulation, testing, and diagnosis. But one bottleneck often prevents the introduction of these ideas: Manual modeling is a non-trivial, time-consuming task. Automatically deriving models by observing and analyzing running systems is one possible way to amend this bottleneck. To derive a model automatically, some a-priori knowledge about the model structure–i.e. about the system–must exist. Such a model formalism would be used as follows: (i) By observing the network traffic, a model of the long-term system behavior could be generated automatically, (ii) Test vectors can be generated from the model, (iii) While the system is running, the model could be used to diagnose non-normal system behavior. The main contribution of this paper is the introduction of a model formalism called 'probabilistic regression automaton' suitable for the tasks mentioned above. %P 177 - 185