A Model for a Testbed for Evaluating Reputation Systems

Abstract

The lack of an universal model in reputation systems makes it challenging to evaluate and compare them against attacks. While there are test beds that provide application domain specific metrics to evaluate reputation systems, in this paper we propose a model for a test bed that is application agnostic. It is a workflow of graph transformations that is generic enough to accommodate a number of reputation systems in existing literature. In doing so, we note that these reputation systems work at different stages in the workflow and as a result, a byproduct of this model is a new classification method. We also describe various attacks using our model.