The document proposes solutions for approximate k-nearest neighbor queries that preserve both the location and query privacy of mobile users. The basic solution allows a user to retrieve one type of point of interest, such as the approximate k nearest car parks, without revealing to the location-based service provider what type of point is retrieved. The generic solution can be applied to queries with multiple discrete attribute types. Compared to existing solutions, the proposed solutions provide more efficient approximate k-nearest neighbor search while protecting user privacy. Experiments show the basic solution is practical for real-world queries.