A privacy-enhancing model for location-based personalized recommendations

To receive personalized recommendation, users of a location-based service (e.g., a Location-Based Social Network, LBSN) have to provide personal information and preferences to the location-based service. However, detailed personal information could be used to identify the users, and hence compromise user privacy. In this paper, we consider an untrusted third party recommendation service used by the location-based service that may attempt to identify the sender of a recommendation query from the query log or may publish the query log. To protect user identity, anonymization must be done “online” before a query reaches the recommendation service. This is different from the usual “offline” scenario where a trusted recommendation service will receive all unanonymized queries and the focus is to anonymize the collected query log. We propose the notion of online anonymity to formalize this online requirement. The challenge for providing online anonymity is dealing with unknown and dynamic location-based service users who can get online and offline at any time. We define this problem, discuss its implications and differences from the problems in the literature, and propose a solution. Our experimental study shows that it is feasible to achieve personalized recommendation while preserve user privacy.