Workload modelling for mix-based anonymity services

Abstract

Evaluating and improving the performance of mix-based anonymity systems in a real-world setting is critical to foster their adoption. However, current research in this field mostly employs unrealistic models for evaluation purposes. Moreover, previously documented results are often difficult to reproduce. We propose two complementary models tailored to the evaluation of mix-based anonymity services. The models enable realistic experiments and are easy to use as they allow to automatically extract workloads from trace files recorded in real networks and replay them in simulations. We also describe our ready-to-use open source evaluation suite that implements the models. Given the suite, researchers can easily create and re-use well-defined workload sets for evaluation purposes. The workloads can be replayed both in discrete-event simulations and distributed experiments. With this initiative we want to foster open research in our discipline.