Trust in Crowds: Probabilistic Behaviour in Anonymity Protocols

Abstract

Abstract. The existing analysis of the Crowds anonymity protocol assumes that a participating member is either 'honest' or 'corrupted.' This paper generalises this analysis so that each member is assumed to maliciously disclose the identity of other nodes with a probability determined by her vulnerability to corruption. Within this model, the trust in a principal is defined to be the probability that she behaves honestly. We investigate the effect of such a probabilistic behaviour on the anonymity of the principals participating in the protocol, and formulate the necessary conditions to achieve 'probable innocence.' Using these conditions, we propose a generalised Crowds-Trust protocol which uses trust information to achieves 'probable innocence' for principals exhibiting probabilistic behaviour.

Towards a Comprehensive Analysis of Crowds Anonymity System

“…The existing work on the analysis of Crowds anonymity most simply divide the users for honest or corruption, and the authors of [3] think that is not practical and trust plays a major role in deciding the forwarder. With this work the paper aims to make the next step in the security analysis of Crowds.…”

“…In contrast, an anonymous e-mail system may require very strong anonymous (for example, beyond suspicion), without any restriction of bandwidth or latency. 3 A [28] is designed to meet specific performance and security for applications. Its designers believe that the existing anonymity scheme in order to pursue anonymity expense to much performance.…”

“…Its designers believe that the existing anonymity scheme in order to pursue anonymity expense to much performance. 3 A proposes a routing algorithm that takes into account performance and anonymity. By adjusting the parameters, users can choose a suitable path to meet their need.…”

Trust in Anonymity Networks

“…In a recent paper of ours [30] we have analysed the anonymity provided by Crowds extended with some trust information, yet against a completely different threat model. The two papers differ in several ways.…”

“…The two papers differ in several ways. Firstly, [30] considers a global and 'credentialbased' trust notion, unlike the individual-and-reputation-based trust considered here. Secondly, in [30] we considered an attack scenario where all protocol members are honest but vulnerable to being corrupted by an external attacker.…”