Evaluation of P2P Search Algorithms for Discovering Trust Paths

摘要

Distributed security models based on a 'web of trust' eliminate single points of failure and alleviate performance bottlenecks. However, such distributed approaches rely on the ability to find trust paths between participants, which introduces performance overhead. It is therefore of importance to develop trust path discovery algorithms that minimize such overhead. Since peer-to-peer (P2P) networks share various characteristics with the web of trust, P2P search algorithms can potentially be exploited to find trust paths. In this paper we systematically evaluate the application of P2P search algorithms to the trust path discovery problem. We consider the number of iterations required (as expressed by the TTL parameter) as well as the messaging overhead, for discovery of single as well as multiple trust paths. Since trust path discovery does not allow for resource replication (usual in P2P applications), we observe that trust path discovery is very sensitive to parameter choices in selective forwarding algorithms (such as K-walker), but is relatively fast when the underlying network topology is scale-free.