Research On Privacy Preserving Publishing Of Big Location Data Based On Differential Privacy

The widespread popularit y of big data technology has accelerated the pace of mankind's entry into the data age.Although various big data services based on locat ion informat ion provide users with personalized services such as clothing,food,housing,and transportat ion,they also lead to the risk of personal privacy leakage.The stat istical informat ion release of big locat ion data is based on the stat ist ics in the part it ion regions.How to improve the qualit y of publishing service of big locat io n data stat ist ical informat ion release is an urgent problem under the premise of ensuring the privacy of user's location.Combining the differential privacy model with the publishing of big locat ion data stat istics can realize the privacy protect ion of big locat ion data publishing informat ion under any background knowledge.The reasonable allocat ion of the privacy budget in the different ial privacy model is related to the availabilit y of the stat ist ical informat ion release of big locat ion data,a hierarchical incremental allo cation method of different ial privacy budget is proposed.According to the commo n tree hierarchical structure,from root node to leaf node,the privacy budget allocat ion is increased layer by layer according to the method of increasing tolerance,which reduces the error of region count ing query.In view of some asymmetric hierarchical tree partit ion structures,corresponding privacy budget adjust ment method is proposed to ensure the combined characterist ics of different ial privacy.Experiments and analysis on large data sets of actual locat ions show that the different ial privacy budget hierarchical incremental allocat ion method proposed in this thesis is superior to other exist ing privacy budget allocat ion methods in improving the query accuracy of published data area.The part it ion structure is very important for the stat ist ical informat ion release o f big locat ion data.The traditional tree partit ion structure is difficult to give reasonable part it ion stopping condit io ns,and the distributio n state of locat ion data in space is not fully considered,which makes the noise error and uniform assumpt ion error of part it ion publishing relat ively large,which reduces the query accuracy of published data.This thesis designs an unbalanced quadtree partit ion method based on the regional uniformit y as a judgement condit io n.According to the spat ial distribut io n densit y of locat ion data,it adapt ively part it ions the quadt ree iterat ively,reducing the over-part it ioning of node sparse regions on the published data.By means of the privacy budget adjust ment method,the privacy budget combinat ion characterist ics under the unbalanced quadtree part it ion structure are realized.T he experimental results show that the unbalanced quadtree partit io n method proposed in this thesis has significant ly improved the availabilit y of published data.