Abstract
A location-based recommender system (LbRS) is a system which provides recommendations to a user related to his/her point of interest. In order to generate these recommendations, the LbRS uses personal information regarding the current location of the user. This creates serious privacy issues, as users' movements can be revealed through their location data. This paper proposes a new location-based privacy protection method, and is divided into two stages. First, dummy locations are identified using query probability and distance. Second, a deep learning algorithm is trained to predict dummy locations. Then, the average entropy of each stage is used to compute final entropy. The results show that the proposed method outperforms standard methods such as random and farthest dummy location selection, although it is fractionally slower than the benchmark methods due to the encryption mechanism integrated into it to provide double-layer security.