Pervasive computing applications typically involve cooperation among a number of entities spanning multiple organizations. Any security breach in any single entity can have very far-reaching consequences. In addition, a number of factors make the task of defending against malicious attacks in pervasive systems even more complex than conventional systems. Foremost among them is that a significant number of the devices deployed in such environments are frequently severely resource constrained. Thus strong security controls cannot be easily deployed on these devices. A second factor is that since a large number of such devices are also involved, attacks can propagate very fast in pervasive environments. These prompt us to propose a model for predicting malicious activities in pervasive systems. Our model is based on a logic of opinion that has been proposed elsewhere. Ours is not an intrusion detection system for pervasive systems but works in tandem with one. The system we propose can be...