Social Networks has rapidly become a fundamental player of today's Internet and have changed the way users generates content and interacts with each other. Recent events have shown the importances of these social medias and point out some drawbacks due to its centralized infrastructure. Privacy concerns, data ownership, resistance to censorships and infrastructure failures give rise to many questions. To address these new challenges, the idea to switch from centralized system to a peer-to-peer infrastructure for social networks has been proposed. Acquaintances in social networks are mainly explicit, a user choice which other users he want as friends in order to share messages and contents. In contrast to client-server infrastructure, in a decentralized system, the model to build acquaintances (topology) greatly influences the performance of the system. In this context and according to the application, how to create the best neighborhood is an important question. In this paper, we address this question via rumor spreading through decentralized social network. The drawbacks of explicit acquaintances are point out and the benefits of implicit one in term of messages are presented.