In this paper we address the query routing problem in peerto-peer (P2P) information retrieval. Our system builds up on the idea of a Semantic Overlay Network (SON), in which each peer becomes neighbor of a small number of peers, chosen among those that are most similar to it. Peers in the network are represented by a statistical Language Model derived from their local data collections but, instead of using the non-metric Kullback-Leibler divergence to compute the similarity between them, we use a symmetrized and "metricized" related measure, the square root of the JensenShannon divergence, which let us map the problem to a metric search problem. The search strategy exploits the triangular inequality to efficiently prune the search space and relies on a priority queue to visit the most promising peers first. To keep communications costs low and to perform an efficient comparison between Language Models, we devise a compression technique that builds on Bloom-filters and histograms and we provide error bounds for the approximation and a cost analysis for the algorithms used to build and maintain the SON.

Semantic overlay networks are an instance of unstructured overlays, where System OverviewSemantic Overlay Networks (SONs) have proven a useful technology not only for distributed information retrieval, but also as a natural distributed alternative to Web 2.0 application domains such as decentralized social networking in the spirit of Flickr or del.icio.us. In this work we present Clouds, a novel system that provides anonymity and censorship resistance over SONs. Anonymity is achieved by relying on a self-organization of peers into groups that we call clouds. Message routing takes place among clouds instead of peers, thus hiding the identity of both the resource provider and the querying peer, while cloud size is a tunable parameter that affects anonymity and efficiency. Censorship resistance at communication level is achieved by a cryptographic protocol that guarantees the secrecy of the resource, thus avoiding censorship based on the inspection of the messages circulating in the network. The design of such a protocol needs to meet a number of challenging goals: allowing for the exchange of encrypted messages without assuming previously shared secrets, avoiding centralized infrastructures, like trusted servers or static gateways (e.g., as in [2]), and guaranteeing efficiency without establishing direct connections between peers. Clouds is the first system to guarantee anonymity and censorship resistance in SONs by addressing the aforementioned challenges.Anonymity in Clouds is achieved by cloaking both the querying peer and the resource provider behind a group of neighboring peers, called cloud. Peers generate clouds at random, without necessarily using them, to minimize the correlation between the events of joining and using a cloud. They also non-deterministically Copyright is held by the author/owner(s). PODC'08, August 18-21, 2008, Toronto, Ontario, Canada. ACM 978-1-59593-989-0/08/08. Figure 1. Overview of Clouds protocol.decide to participate or not in clouds created by other peers. By design, clouds are populated by peers in the neighborhood of the cloud initiator. Communication takes place between clouds, and all peers in a cloud share the same probability of being involved in any communication which has this cloud as the start-or endpoint (k-anonymity). Clouds protocol is designed so that the observable behavior is the same for all peers, regardless of initiating of forwarding a message, to avoid compromising anonymity. The proposed cryptographic protocol aims at addressing the problem of censorship at the communication level, where a malicious party aims at filtering out any communication (i.e., query or resource) that contains unwanted content. The privacy of the resource is protected by cryptography, making it hard for the attacker to censor the communication by inspecting the message content.The protocol is composed of four steps summarized in Figure 1. A querying peer P uses a cloud C P it participates in to issue a query q. This query follows a random walk in the cloud to obscure the me...

Abstract. This paper presents Clouds, a peer-to-peer protocol that guarantees both anonymity and censorship resistance in semantic overlay networks. The design of such a protocol needs to meet a number of challenging goals: enabling the exchange of encrypted messages without assuming previously shared secrets, avoiding centralised infrastructures, like trusted servers or gateways, and guaranteeing efficiency without establishing direct connections between peers. Anonymity is achieved by cloaking the identity of protocol participants behind groups of semantically close peers. Censorship resistance is guaranteed by a cryptographic protocol securing the anonymous communication between the querying peer and the resource provider. Although we instantiate our technique on semantic overlay networks to exploit their retrieval capabilities, our framework is general and can be applied to any unstructured overlay network. Experimental results demonstrate the security properties of Clouds under different attacks and show the message overhead and retrieval effectiveness of the protocol.

No abstract