Xiuqi Li

Cluster-based intelligent searching in unstructured peer-to-peer networks

Existing cluster-based searching schemes in unstructured peer-to-peer (P2P) networks employ flooding/random forwarding on connected dominating sets (CDS) of networks. There exists no upper bound on the size of CDS of a network. Both flooding and CDS hinder query efficiency. Random forwarding worsens the recall ratio. In this paper, we propose a cluster-based searching scheme that intelligently forward queries on the maximum independent sets (MIS) of networks. Our approach partitions the entire network into disjoint clusters with one clusterhead (CH) per cluster. CHs form a MIS and are connected through gateway nodes. Each node fakes one role, a CH, a gateway, or an ordinary node. A CH looks up the data for the entire cluster using data summaries of cluster members, which are represented by bloom filters. Between clusters, CHs intelligently forward queries via gateways to the best neighbor CHs that are most likely to return query results. The experimental results demonstrate that our scheme greatly improves the query efficiency without degrading the quality of the query results, compared to existing approaches.

A Class-Based Search System in Unstructured P2P Networks

Efficient searching is one of the important design issues in peer-to-peer (P2P) networks. Among various searching techniques, semantic-based searching has drawn significant attention recently. Gnutella-like efficient searching system (GES) in the work of Zhu et al. (2005) is such a system. GES derives a node vector, a semantic summary of all of the documents on a node, based on vector space model (VSM). The topology adaptation algorithm and search protocol are then designed according to the similarity between node vectors of different nodes. However, although GES is suitable when the distribution of documents in each node is uniform, it may not be efficient when the distribution is diverse. When there are many categories of documents at each node, the node vector representation may be inaccurate. We extend the idea of GES and present a class-based semantic searching system (CSS). It makes use of a data clustering algorithm, online spherical k-means clustering (OSKM) in the work of Zhang (2005), to cluster all documents on a node into several classes. Each class can be viewed as a virtual node. Virtual nodes are connected through virtual links. As a result, class vector replaces node vector and plays an important role in the class-based topology adaptation and search process, which makes CSS very efficient. Our simulation using the IR benchmark TREC collection demonstrates that CSS outperforms GES in terms of higher recall, higher precision and lower search cost.

Hint-based routing in WSNs using scope decay bloom filters

In existing query-based routing protocols in wireless sensor networks (WSNs), a node either keeps precise route information to desired events, such as in event flooding, or does not keep any route to desired events such as in query flooding. In this paper, we propose a routing protocol, called Hint-based Routing by Scope Decay Bloom Filter (HR-SDBF), that employs probabilistic hints. In HR-SDBF, each node maintains some probabilistic hints about events and utilizes these hints to route queries intelligently. We also put forward a data structure, scope decay bloom filter (SDBF) to encode the probabilistic hints. With SDBF, the amount of information about an event is propagated, without any loss, within the k-hop neighborhood of an event source but decreases outside the k-hop neighborhood as the distance from the event source increases. Compared to existing query-based protocols, HR-SDBF greatly reduces the amortized network traffic without compromising the query success rate and achieves a higher energy efficiency. To the best of our knowledge, this is the first query routing protocol in WSNs that utilizes probabilistic hints encoded in a variant of the bloom filter. Both the analytic and the experimental results support the performance improvement of our protocol

Improve Searching by Reinforcement Learning in Unstructured P2Ps

Existing searching schemes in unstructured P2Ps can be categorized as either blind or informed. The quality of query results in blind schemes is low. Informed schemes use simple heuristics that lack the theoretical background to support the simulation results. In this paper, we propose to improve searching by reinforcement learning (RL), which has been proven in artificial intelligence to be able to learn the best sequence of actions in order to achieve a certain goal. Our approach, ISRL (intelligent searching by reinforcement learning), aims at locating the best path to desired files at low cost. It explores new paths by forwarding queries to randomly chosen neighbors. It also exploits the paths that have been discovered to reduce the cumulative query cost. Two models of ISRL are proposed: the basic ISRL for finding one desired file, and MP-ISRL (multipath ISRL) for finding multiple desired files. ISRL outperforms existing searching approaches in unstructured P2Ps by achieving higher query quality with less query traffic. The experimental result supports the performance improvement of ISRL.

Improve Peer Cooperation Using Social Networks

Due to the dynamic nature of P2P systems, it is impossible to keep an accurate history of the transactions that take place while avoiding security attacks such as whitewashing and collusion, and abuse such as freeriding. This is why it is important to develop a mechanism that both rewards cooperative peers and punishes misbehaving peers. Modelling P2P networks as social structures can allow incentive mechanisms to be developed that prevent the negative behaviors mentioned. In a social structure, peers make and receive payments for services provided to and from each other. In this paper we extend a social network algorithm to include the transfer of credit between peers to reduce the path length in queries. We also develop a selection strategy that involves different aspects of peer interactions in a P2P network and a credit transfer mechanism that helps to dis-incent misbehaving peers by taking away credits that they have with good peers and transferring them to more cooperative ones. The simulation results show that our algorithm is effective in reducing the amount of debt between peers, meaning that peers become more cooperative, and shortening the average path length to a satisfied query, while increasing delivery ratio.

HR-SDBF: an approach to data-centric routing in WSNs

In existing query-based routing protocols in wireless sensor networks (WSNs), a node either keeps precise route information to desired events, such as in event flooding, or does not keep any route to desired events, such as in query flooding. In this paper, we propose a routing protocol, called hint-based routing by scope decay bloom filter (HR-SDBF), that employs probabilistic hints. In the HR-SDBF protocol, each node maintains some probabilistic hints about the potential desired events and routes queries intelligently based on these probabilistic hints. We also put forward a data structure, scope decay bloom filter (SDBF) to encode the probabilistic hints. With SDBF, the amount of information about an event is propagated, without any loss, within the k-hop neighbourhood of the event source, but decreases outside the k-hop neighbourhood as the distance from the event source increases. Compared to existing query-based protocols, HR-SDBF greatly reduces the amortised network traffic without compromising the query success rate, and achieves a higher energy efficiency. To the best of our knowledge, this is the first query routing protocol in WSNs that utilises probabilistic hints encoded in a variant of the bloom filter. Both the analytic and the experimental results support the performance improvement of our protocol.

Improve peer cooperation using social networks

Due to the dynamic nature of P2P systems, it is impossible to keep an accurate history of the transactions that take place while avoiding security attacks such as whitewashing and collusion, and abuse such as freeriding. This is why it is important to develop a mechanism that both rewards cooperative peers and punishes misbehaving peers. Modelling P2P networks as social structures can allow incentive mechanisms to be developed that prevent the negative behaviors mentioned. In a social structure, peers make and receive payments for services provided to and from each other. In this paper we extend a social network algorithm to include the transfer of credit between peers to reduce the path length in queries. We also develop a selection strategy that involves different aspects of peer interactions in a P2P network and a credit transfer mechanism that helps to dis-incent misbehaving peers by taking away credits that they have with good peers and transferring them to more cooperative ones. The simulation results show that our algorithm is effective in reducing the amount of debt between peers, meaning that peers become more cooperative, and shortening the average path length to a satisfied query, while increasing delivery ratio.

ISRL: intelligent search by reinforcement learning in unstructured peer-to-peer networks

Existing searches in unstructured peer-to-peer (P2P) networks are either blind or informed based on simple heuristics. Blind schemes suffer from low query quality. Simple heuristics lack theoretical background to support the simulation results. In this paper, we propose an intelligent searching scheme, called intelligent search by reinforcement learning (ISRL), which systematically seeks the best route to desired files by reinforcement learning (RL). In artificial intelligence, RL has been proven to be able to learn the best sequence of actions to achieve a certain goal. To discover the best path to desired files, ISRL not only explores new paths by forwarding queries to randomly chosen neighbors, but also exploits the paths that have been discovered for reducing the cumulative query cost. We design three models of ISRL: a basic version for finding one desired file, MP-ISRL for finding at least k files, and C-ISRL for reducing maintenance overhead through clustering when there are many queries. ISRL outperforms existing searching approaches in unstructured P2P networks by achieving similar query quality with lower cumulative query cost. The experimental result confirms the performance improvement of ISRL.