Discussing Anonymity Metrics for Mix Based Anonymity Approaches

Today the Chaumian Mix idea is not only an academic approach for anonymous communication, it is also a practical system with many variants used in real-world scenarios. It is therefore important to find an appropriate measure for the anonymity provided by these approaches. Many measurement approaches have been proposed that consider only the static state of the system without accounting for past and future information. Still other measurements evaluate only statistics. These measurements have in common that they do not measure when the anonymity function of the system is broken. Inspired by the idea of unicity distance in cryptography, and the mean time to failure in dependable systems, we believe that measuring the point at which the system fails to hide the relation between a sender and a receiver is a more appropriate measure of its anonymity. In this paper, we discuss our arguments with respect to existing measurement approaches.