Discovering the Most Trusted Agents without Central Control

The ability to determine the trust of an agent relative to other agents can be useful for applications that require to select a subset of trusted agents. However, in an open, distributed system, it is difficult to realize this functionality. Selecting agents on the basis of trust can be seen as an extension of the usual Trust Management service. In this paper, we propose a fully distributed method for selection of the most trusted agents. We investigate the sensitivity of this method to adversaries and propose improvements to its security. The performance of the method and the quality of the selection is studied in detail.