Résumé en anglais

Service discovery has been adopted in distributed environments and recently in social networks. Existing approaches for service discovery are often based on matching techniques, which only capture common quality of service (QoS) criteria while ignoring the contribution of the social aspect. The discovery problem increases when the number of available services is important and no means to distinguish between two or many providers offering the same service. To overcome this issue, we propose a trust measure defined as the combination of two dimensions namely trust in sociability and trust in expertise. The problem is growing when no central control can be fixed due to the distributed nature of social networks. To cope with centrality related issues, we propose an agent-based service discovery approach in which each user in the social network is assigned an autonomous and non malicious agent equipped with a bounded set of services with their advertised QoS values. Endowed with a limited view of the social network, the agent acts locally on behalf of its user to discover trustworthy providers with good services. The search propagation process within the social network is ensured by means of a distributed referral system where agents cooperate and evaluate referrals based on a distributed knowledge and a decentralized decision-making.