A Testbed for Comparing Trust Computation Algorithms

Trust is the expectation of a person about another person’s behavior. Trust is important for many security related decisions about, e.g., granting or revoking privileges, controlling access to sensitive resources and information, or evaluating intelligence gathered from multiple sources. More often than not, the issue is complicated even further because the person making the decision has no direct trust relationship with every single subject whose trustworthiness needs to be evaluated. So, the decision maker needs to rely on recommendations by others, and then somehow aggregate the trust related information that is collected. In this work we provide an algebraic framework in which we can describe multiple ways that trust related information can be aggregated to form a single value. We show the similarities and differences that the various so called trust computation algorithms have, and associate these with the algebraic properties of the framework that we consider.