Applying Dynamic Bayesian Networks for Automated Modeling in ArchiMate: A Realization Study

Abstract

Enterprise Architecture modeling is an approach to manage modern IT infrastructure and landscapes to coordinate a multitude of IT projects in an organization. Enterprise architects apply modeling tools such as ArchiMate to document the enterprise architecture. Because these models have traditionally been created and maintained manually, efforts to manage IT architecture have been both time-consuming and error-prone. We evaluate an approach by Johnson et al. (2016) for automated generation of these models from observed network traffic using Dynamic Bayesian Networks. As inference in large Dynamic Bayesian Network proves computationally infeasible, we propose an alternative approach using a set of Hidden Markov Models to model the current network state, present an implementation, and evaluate its performance in a real-world setting.