The reverse statistical disclosure attack

Abstract. Statistical disclosure is a well-studied technique that an at-tacker can use to uncover relations between users in mix-based anonymity systems. Prior work has focused on nding the receivers to whom a given targeted user sends. In this paper, we investigate the eectiveness of sta-tistical disclosure in nding all of a users ' contacts, including those from whom she receives messages. To this end, we propose a new attack called the Reverse Statistical Disclosure Attack (RSDA). RSDA uses observa-tions of all users sending patterns to estimate both the targeted user's sending pattern and her receiving pattern. The estimated patterns are combined to nd a set of the targeted user's most likely contacts. We study the performance of RSDA in simulation using dierent mix net-work congurations and also study the eectiveness of cover trac as a countermeasure. Our results show that that RSDA outperforms the tra-ditional SDA in nding the user's contacts, particularly as the amounts of user trac and cover trac rise.