A multigraph approach for supporting computer network monitoring systems

The pervasiveness of information technologies has reached very high levels: most human activities involve the use of sensor-based systems connected to the network. The increasingly widespread use of the Internet of things has significantly improved our quality of life but has introduced a series of new problems, especially from the security point of view. Protecting these systems from cyber-attacks has become a priority as possible malfunctions can lead to issues with a significant social impact. Imagine, for example, computer attacks on smart cars connected to the network or remotely controlled electrical or water systems. Protecting this type of system is a complex task as there are many elements to consider and the data to be monitored. An analysis able to foresee eventual attacks through the study of the data and their variations could be a useful tool to prevent malfunctions. This paper proposes a methodology based on the integrated use of three graphic models to address the problem of preventing attacks on pervasive systems from three different perspectives: probabilistic, contextual, and ontological. The paper proposes the use of Bayesian networks built through an ontological definition of the problem dropped on a particular context represented by a Context Dimension Tree—the proposed approach experiments in a real scenario providing satisfactory results