Range Optimal Dummy Location Selection Based on Query Probability Density

Location semantics are helpful for some location-based services in urban life, such as location retrieval and recommendation. The location semantics in these services may include a very rich and diverse range of location tags and comments. In addition, when using dummy location selection to protect location privacy, if a user initiates a service request in a location with a zero-query probability, the generated location set will only have a zero-query probability in the dummy location set. Attackers can easily obtain the user's true location. Most dummy location selection algorithms do not fully consider the uneven probability of historical queries and richer location semantics. To address this issue, a range optimal dummy location selection algorithm based on query probability density is proposed. This method first uses kernel density estimation to find a replacement location within an acceptable loss range for the user's actual request location. Then, word vectors and cosine similarity are used to describe the semantic distance of the location. Finally, a quadrant dichotomy location selection method is used to generate a dummy location set. Conduct simulation experiments on public datasets, compare with existing methods in terms of location entropy, dummy location set coverage area, minimum physical distance, query probability uniformity, dummy location set generation time, and θ- The comparison of the six aspects of safety values proves the effectiveness of the proposed method.