The federation of different data sources gained increasing attention due to the continuously growing amount of data. But the more data are available from heterogeneous sources, the higher the risk is of inconsistency. To tackle this challenge in federated knowledge bases we propose a fully automated approach for computing trust values at different levels of granularity. Gathering both the conflict graph and statistical evidence generated by inconsistency detection and resolution, we create a Markov network to facilitate the application of Gibbs sampling to compute a probability for each conflicting assertion. Based on which, trust values for each integrated data source and its respective signature elements are computed. We evaluate our approach on a large distributed dataset from the domain of library science.