Measuring anonymity using network coordinate systems

Popularity and awareness of anonymisation systems increased tremendously over the past years, however only a very few systems made it from research to production. These systems usually add intermediate nodes in the communication path aiming to hide user identities. Several attacks against these systems exist, like timing attacks or exploitation of latency information. In this paper, we propose an alternative approach to disclose users of current popular anonymisation systems in practice by the means of virtual network coordinate systems, a widely accepted method for latency prediction and network optimisation. Mapping physical nodes to a n-dimensional space can reveal a geographical proximity that is used to disclose users, who expect to stay anonymous. We define a model that leverages network coordinates in order to measure quantitatively the anonymity services and evaluate it on the Planet-Lab research network. The basic idea is to analyse the relative distance between nodes and to calculate the probability of nodes being hosted in the same location. Evaluation proves that our proposed model can be used as a measure of anonymity.