A novel service discovery model for decentralised online social networks.

Online social networks (OSNs) have become the most popular Internet application that attracts billions of users to share information, disseminate opinions and interact with others in the online society. The unprecedented growing popularity of OSNs naturally makes using social network services as a pervasive phenomenon in our daily life. The majority of OSNs service providers adopts a centralised architecture because of its management simplicity and content controllability. However, the centralised architecture for large-scale OSNs applications incurs costly deployment of computing infrastructures and suffers performance bottleneck. Moreover, the centralised architecture has two major shortcomings: the single point failure problem and the lack of privacy, which challenges the uninterrupted service provision and raises serious privacy concerns. This thesis proposes a decentralised approach based on peer-to-peer (P2P) networks as an alternative to the traditional centralised architecture. Firstly, a self-organised architecture with self-sustaining social network adaptation has been designed to support decentralised topology maintenance. This self-organised architecture exhibits small-world characteristics with short average path length and large average clustering coefficient to support efficient information exchange. Based on this self-organised architecture, a novel decentralised service discovery model has been developed to achieve a semantic-aware and interest-aware query routing in the P2P social network. The proposed model encompasses a service matchmaking module to capture the hidden semantic information for query-service matching and a homophily-based query processing module to characterise user’s common social status and interests for personalised query routing. Furthermore, in order to optimise the efficiency of service discovery, a swarm intelligence inspired algorithm has been designed to reduce the query routing overhead. This algorithm employs an adaptive forwarding strategy that can adapt to various social network structures and achieves promising search performance with low redundant query overhead in dynamic environments. Finally, a configurable software simulator is implemented to simulate complex networks and to evaluate the proposed service discovery model. Extensive experiments have been conducted through simulations, and the obtained results have demonstrated the efficiency and effectiveness of the proposed model.