On the Sender Cover Traffic Countermeasure Against an Improved Statistical Disclosure Attack

The statistical disclosure attack against a particular user of an anonymity system is known to be very effective in determining, after long-term observation of the system, the set of receivers that user sends messages to. This paper first presents an improvement over this attack that, by employing a weighted mean of the observed relative receiver popularity, is more accurate than the original one based upon arithmetic mean. Second, a mathematical analysis is presented of this attack on a model, in which senders blend dummy messages with real ones. It is shown that despite such sender-generated dummy cover traffic, the attack can proceed almost unhindered. The analysis substantiates earlier empirical indications of the ineffectiveness of this countermeasure.