In a reputation-based trust management system an entity's behaviour determines its reputation which in turn affects its interaction with other entities. We present a mathematical model for trust aimed at global computing environments which, as opposed to many traditional trust management systems, supports the dynamics of reputation-based systems in the sense that trusting relationships are monitored and changes over time, depending on the behaviour of the entities involved. The main contribution is the discovery that the notion of event structures, well studied e.g. in the theory of concurrency, can faithfully model the important concepts of observation and outcome of interactions. In this setting observations are events and an outcome of an interaction is a maximal set of consistent events describing what happened. We also touch upon the problem of transferring trust or behavioural information between contexts, and we propose a generalised definition of morphism of event structures as an information-transfer function.
[Trust, Risk, Reputation, Dempster-Shafer Theory of Evidence, Formal Models, Event Structures, Uncertain Probabilites]