Location privacy and query content privacy are both critical elements in LBS querying for points of interest (POIs). For continuous queries in road networks, frequent changes of a user's location bring huge burden of query processing to LBS server, how to release a user's privacy information as little as possible, and obtain accurate query results efficiently are still great challenges in current researches. Taking the idea of private information retrieval (PIR), i.e. no trusted entities except the user himself, as a basic assumption, a privacy-preserving method is put forward based on homomorphic properties of Paillier cryptosystem, which the user does not need to provide his actual location or query content to LBS server in K nearest neighbor POIs query, it achieves privacy preservation in LBS and accurate retrieval of POIs. Meanwhile, takeing the vertexes in road networks as generating elements to organize the distribution information of POIs, the inefficient problem is further solved in most cryptographic query schemes, which is caused by frequent location changes in continuous query, the proposed method significantly reduces the frequency of initiating queries to LBS server without decreasing the query accuracy. Finally, the proposed method is analyzed from the aspects of accuracy, security, and efficiency, extensive experiments verify the effectiveness.