Data releasing is a key part bridging between the collection of big data and their applications. Traditional methods release the static version of dataset or publish the snapshot with a fixed sampling interval, which cannot meet the dynamic query requirements and query precision for big data. Moreover, the quality of published data cannot reflect the characteristics of the dynamic changes of big data, which often leads to subsequent data analysis and mining errors. This paper proposes an adaptive sampling mechanism and privacy protection method for the release of big location data. In order to reflect the dynamic change of data in time, we design an adaptive sampling mechanism based on the proportional-integral-derivative (PID) controller according to the temporal and spatial correlation of the location data. To ensure the privacy of published data, we propose a heuristic quad-tree partitioning method as well as a corresponding privacy budget allocation strategy. Experiments and analysis prove that the adaptive sampling mechanism proposed in this paper can effectively track the trend of dynamic changes of data, and the designed differential privacy method can improve the accuracy of counting query and enhance the availability of published data under the premise of certain privacy intensity. The proposed methods can also be readily extended to other areas of big data release applications.
Bibliographical noteVersion archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.
- adaptive sampling
- Big location data
- differential privacy
- heuristic quad-tree partitioning
- privacy preserving data publishing