Enabling attack behavior prediction in ubiquitous environments

The pervasive computing paradigm has raised issues such as conceptual semantic descriptions and ambient management of information resources. The probabilistic theory on the other hand provides uncertain knowledge representation schemes that are semantically inefficient. However, security models related to attacks exploits both semantic and probabilistic modeling. Issues such as attack prediction and classification of attacker's intentions are of high importance in IDS environments. In this paper we propose a novel Breadth and Depth Bayesian classifier and an inference probabilistic algorithm. The inference algorithm is applied over well defined conceptual information integrated in a hybrid IDS by means of ontologies.