Passive-attack analysis for connection-based anonymity systems

Abstract

In this paper we consider low-latency connection-based anonymity systems which can be used for applications like web browsing or SSH. Although several such systems have been designed and built, their anonymity has so far not been adequately evaluated.We analyse the anonymity of connection-based systems against global passive adversaries. We give a precise description of a packet-counting attack which requires a very low degree of precision from the adversary, evaluate its effectiveness against connection-based systems depending on their size, architecture and configuration, and calculate the amount of traffic necessary to provide a minimum degree of protection. We then present a second attack based on tracking connection starts which gives us another lower bound on traffic volumes required to provide at least some anonymity.

A flexible in-network IP anonymization service

“…While they are still considered "low-latency" connection-oriented approaches compared to slower message-based anonymity systems [20], they still exhibit non-trivial overhead and noticeable delay (as observed in Figure 1. In particular, a recent usability study [6] of Tor found that DNS requests were 40 times slower than direct connections.…”

Privacy of encrypted voice-over-IP

“…Serjantov et al [8], for example, proposes the use of "spikes" in the traffic to find the communication relationship between users. Fu al.…”

“…Many attacks against MIXes and networks of MIXes exist. For example, Serjantov and Sewell [8] analyzed the possibility of a lone flow on an input link of a MIX. If the rate of this lone input flow is approximately equal to the rate of a flow out of the MIX, this pair of input and outflow flows are correlated.…”

Survey on Anonymity in Unstructured Peer-to-Peer Systems

“…B. N. Levine et al [52] presented a technique to thwart timing attacks in Mix-based systems. A. Serjantov and P. Sewell [53] investigated the time analysis for connection-based systems, e.g., Mix-based systems. X. Fu et al [54,55] proposed several attacking models based on traffic analysis, which are effective to thwart the anonymity in both wired and wireless systems.…”