Research On The Statistical Partition Publishing And Privacy Preserving Method Of Big Location Data

Big location data is widely used in popular areas such as Intelligent Transportation System(ITS),Internet of Things(IoT),Internet of Vehicles(IoV),Location-Based Service(LBS),and is closely related to a large amount of private information such as personal living habits,health status,places of interest,economic conditions and so on.Partition publishing is an effective form of releasing big location data statistical information.By adding random noise that satisfies the differential privacy model,it is possible to achieve privacy protection for the big location data,and then improve the application security of location information.Aiming at the privacy protection issue of the statistical release of big location data,this thesis analyzes the process of partition publishing and the implementation method of privacy protection of big location data.In order to make full use of the spatial-temporal correlation of big location data and improve the availability of published data,a differential privacy partition and publishing method based on sampling and adjustment is proposed.First,the continuous release of big location data snapshots is achieved through the sampling with fixed time intervals.A differential processing method is designed to reduce the temporal and spatial redundancy between adjacent data snapshots,thereby reducing the amount of data needs to be processed and improving the operating efficiency of the data publishing algorithm.Then,the impact of the updated data on the partition structure is determined according to the result of the differential processing.The partition and adjustment methods are designed based on the grid structure and the tree structure respectively.Finally,differential privacy protection of statistical release of big location data is achieved by adding Laplace noise to the adjusted partition structure.Experiments and analyses carried out on synthetic and real-world big datasets show that the proposed adjustment algorithm in this thesis has advantages in improving algorithm operation efficiency and range count query accuracy.In order to make the released data better reflect the characteristics of the dynamic change of big location data and meet the users' dynamic query needs for statistical location information,an adaptive sampling mechanism based on the PID controller and a differential privacy partition release method are proposed.First,the actual data snapshot is obtained by sampling the big location data at uniform time intervals and the predicted data snapshot is obtained by predicting the location big data released at the next moment according to the maximum moving boundary and the minimum moving boundary.Then,the feedback error is obtained by comparing the count values of the predicted data snapshot and the actual data snapshot,the PID controller is guided to adaptively adjust the sampling interval in order to determine a reasonable sampling time.Finally,a heuristic Quad-tree partition method and differential privacy budget allocation and post-adjustment strategy is designed to achieve reasonable partition of two-dimensional regions and the differential privacy protection of published data.Experiments and analyses carried out on the real-world big location datasets show that the proposed adaptive sampling mechanism can effectively track the dynamic changing trend of big location data.The proposed differential privacy budget allocation and postadjustment method improve the precision of range counting queries and the availability of published data while ensuring privacy protection.