Countless Internet applications and platforms make it easy to communicate, to collaborate and to exchange information and opinions with a huge number of individuals. However, it is getting more and more difficult to distinguish honest and reliable individuals from malicious users distributing false information or malware. Digital signatures, webs of trust and reputation systems can help to securely identify communication partners and to estimate the trustworthiness of others, but there is a lack of trust and authenticity evaluation methods that do not show counterintuitive effects in the case of conflicting opinions. This article proposes a new integrated method to evaluate uncertain and conflicting trust and authenticity statements. It introduces a set of operators and inference rules for combining and reasoning with these statements, it defines an approach to compute the resulting confidence values of derived statements and it compares different computation algorithms. The computation is based on a probability theoretical model in order to exclude inconsistencies and counter-intuitive effects.