A novel search scheme is proposed to address the efficiency and scalability issu es in unstructured peer-to-peer networks. Unlike the notorious search scheme of flooding in unstructured peer-to-peer networks, the proposed scheme is based on the domain of interest that is formed by nodes sharing the same interest. Search request is propagated within the domain of interest, and the efficiency of sear ch process is very high after nodes in the same domain of interest get familiar with one another. In the proposed scheme, attributes of document are described b y metadata using RDF statements to support flexible queries; nodes having the sa me metadata belong to the same domain of interest. With the facility of metadata selection window and metadata replication, performance of the proposed search s cheme is further improved. Simulation results prove the effectiveness and efficiency of the proposed scheme.