This work aims to provide a personalized recommender system for item selection in decentralized environments. To do so, the recommendation takes into account users' profiles as well as users' social network preferences (ie friends profiles). Using a local approach, our recommender allows users to control entirely scores and profiles sharing.