Differentially private geospatial data publication based on grid clustering

Abstract

Collecting geospatial data from location-based services can provide location evidence while analysing spatial information. However, releasing location data may result in the disclosure of sensitive personal information. The adaptive grid method (AG) uses differential privacy to protect information. In AG, the algorithm uses two levels of grids over data domain. However, it does not take into account the data distribution. Usually, the accuracy will be reduced in response to long-range counting queries. In this paper, the adjacent grid cells with similar data density are clustered together. Laplace noise is added to the clusters created by the clustering of the grid cells. The noisy count obtained from the grid cells that form each cluster is evenly redistributed to the grid cells in the cluster. Extensive experiments on real-world datasets showed that the query accuracy of the proposed method is higher than the existing methods.