In the current big data era, recommendation systems play an important role in our daily life to help us make faster and better decisions from massive amounts of choices. However, the exposure of sensitive data raises a big privacy concern. Research has shown that it is possible to de-identify anonymous users, i.e., inferring sensitive information from non-sensitive data such as movie ratings). Hence, in this M.Sc. thesis, I tackle the privacy issue by designing a semi-decentralized network called Trust-based Agent Network (TAN). It treats each node in the network as an agent, and it comprises two key concepts. First, data are distributed to each agent inside each trusted network. Second, the recommendation service provider can only collect obfuscated data from agents by adopting the differential privacy mechanism. As a result, in the TAN, data are either protected inside local trusted networks or obfuscated outside of trusted networks. Then, a final recommendation can be made by aggregating the local suggestions from the trusted network and obfuscated global suggestions from the service provider. Note that the resulting recommendation emphasizes the recommendations from the trusted agents. Knowing the potential limitation of the reduced search space (which can cause inaccuracy in the recommendation result), the resulting recommendation also takes into account the general suggestions from the whole network. Experimental evaluation results show that my proposed solution protects privacy while preserving good accuracy.