With the abundance of location-aware portable devices such as cellphones and PDAs, a new emerging application is to use this pervasive computing platform to learn about the whereabouts of one's friends and relatives. However, issues of trust, security and privacy have hindered the popularity and safety of the systems developed for this purpose. We identify and address the key challenges of enabling private spatial queries in social networks using an untrusted server model without compromising users' privacy. We propose private buddy search (PBS), a framework to enable private evaluation of spatial queries predominantly used in social networks, without compromising sensitive information about its users. Utilizing server side encrypted index structures and client side query processing, PBS enjoys both scalability and privacy. Our extensive experimental evaluation shows that PBS supports very efficient user operations such as location updates, as well as spatial queries such as range and k-nearest neighbor search.