Study On Geo-indistinguishability For Multi-positions Based On Mechanism Of Adding Noise To The Centroid Of Positions

The rapid growth of the volume of data in the Big Data Time, resulting in a large number of private data was exposed, privacy protection become a serious problem to be solved, the increase of the location service makes the location privacy protection become even more important. Most existing protection models are unable to protect the attack based on the background knowledge, and do not have a reasonable measurement for the privacy protection level. Differential privacy protection model is targeting with background knowledge model, quantify ing privacy protection level, it is a rigorous model of the privacy protection. In 2013, it is applied to the location privacy protection for the first time by Miguel Andres and others, thus producing the ?-geo-indistinguishability, bringing a new blood to location privacy protection research. The basic idea of ?- geo- indistinguishability is to add random noise which satisfies a specified distribution to the primitive positions, but when the quantity of positions becomes large, using the mechanism of adding random noise independently to the primitive positions results in the problem of inducing too much error.To solve this problem, this paper proposes a noise adding mechanism based on on a single centroid for the ?- geo- indistinguishability of multi-positions, expounded the definitions and noise generating method of the ?- geo- indistinguishability based on a single centroid, and compared the error between this mechanism and independent mechanism. Experimental results show that when the distance between the positions meets a certain threshold, the noise adding mechanism based on a single centroid for the ?-geo- indistinguishability of multi positions performs better than independent mechansim. O n the basis of the noise adding mechanism based on a single centroid, this paper proposes a noise adding mechanism based on multi-centroids for the ?-geo-indistinguishability of multi-positions, the algorithm clusters the positions first, and then add noise to the clusters to reduce the total error. The experiment results show that, under the same privacy protection level, the mechanism based on adding noise to multi-centroids of the positions for the ?-geo- indistinguishability of multi-positions performs better than independent mechanism, inducing less total error.