Active or inactive: infer private user information in location-based social network

Private user information can be compromised while revealing individual location data in widely used location-based social networks (LBSNs). In order to reveal the risk of location privacy faced by users, we demonstrate a method, which transforms social networks into Bayesian networks, to infer private information through the location data and relationships among users in LBSNs, such as Gowalla, regardless of whether users are active or inactive. Location data from active users can be easily used to infer private information like consumption level. For example, people who frequently appear in expensive restaurants are likely to rank the high consumption level. Those inactive users, who share sparse location data, reveal their private information through their active friends whose private information is easily divulged. Our experimental results show that friends have a high probability of having been to the same places. Combining with relationship data, the possibility of revealing private information is dramatically improved.