A trust evaluation method based on logic and probability theory

Abstract

We introduce a trust evaluation method applicable in a decentralized setting, in which no universally trusted authority exists. The method makes simultaneous use of logic and probability theory. The result of the qualitative part of the method are logical arguments for and against the reliability of an entity. The quantitative part returns the probability that the reliability of an entity can be deduced under the given assumptions and pieces of evidence, as well a corresponding probability for the counter-hypothesis. Our method is a true generalization of existing methods, in particular the Credential Networks. It relies on digital signatures for authenticating messages and accounts for many-to-many relationships between entities and public keys. Moreover, it includes eight different types of trust relations, namely the assumption or the statement that an entity is honest, competent, reliable, or malicious, and their corresponding negations. © 2008 International Federation for Information Processing.