Measuring client-server anonymity

Abstract

The primary intent of clients performing anonymous online tasks is to conceal the extent of their already encrypted communication with sensitive servers. We present an accurate method for evaluating the amount of anonymity still available to such clients in the aftermath of an attack. Our method is based upon probabilities arrived at by the attack of possible client-server association levels for being the real one, along with the correctness levels of those associations. We demonstrate how additionally taking correctness levels into account results in more accurate anonymity measurement than the customary approach of just computing the Shannon entropy of the probabilities.