Evaluating the Effectiveness of Anonymity Communication Systems Against Malicious Insiders

The traffic analysis problem is one in which the communication partners of Internet users are learned by unrelated parties. Few works have been devoted to evaluating the effectiveness of the proposed anonymity designs. This paper proposes a unique adversary model to quantitatively measure how much precise information an attacker can obtain while subverting anonymity systems. The results of simulation experiments offer some insights into designs of anonymity protocols: (1) longer routing paths do not necessarily preserve anonymity better; (2) as the attacker controls more nodes, cyclic loops tend to create more confusing effects with overwhelming information; (3) reputation-based route selection may increase the efficiency, as long as the trusted routing nodes are truly trustworthy.