Research On Association Rules And Sequence Patterns Mining From LBS Spatial-Temporal K-Anonymity Datasets

With the development of Location-Based Services system (LBS), people are going to pay moreand more attention to privacy protection of LBS. Then, privacy protection of LBS is becoming ahotspot topic in GIS and mobile communication domains. Although a series of LBS privacyprotection methods have been proposed, they can’t resist attacks based on data mining techniques.In order to verfy this problem, this paper conducted the following research:(1) based on theanalysis of the related literature theory, first proposed the traditional spatio-temporal K-anonymousand its variants may can not effectively defense analysis inference attack problem of large spatialand temporal scales based on anonymous set, for this purpose, the design and development of thesoftware system of LBS space-time K-anonymity experimental data generated from GPS track datasimulation. The research object of this project is a series of anonymous data generated in LBSspace-time K-anonymous query process. But at present Spatial Temporal K-anonymous andoptimization method has not been widely used in the commercial LBS system, so we firstlypreprocess the GPS trajectory data, the experimental simulation of anonymity set data to a certainscale generated from GPS data based on spatio-temporal K-anonymous algorithm proposed byGruteser et al. GPS trajectory data of this experiment, from a taxi management company in Nanjingcity in July15,2007, collection (0：00-24：00)2612taxi GPS trajectory data period. Experimentalsimulation of anonymous data production include： LBS snapshot query anonymity set, LBScontinuous query anonymity set.(2) we designed the methods of mining association rules andsequential patterns from LBS Spatial Temporal K-Anonymity datasets,then use the twomethods,with anonymity set snapshot data generated from a simulated GPS data and the anonymityset sequence as experimental data,respectively for the mining experiment anonymity set associationrules and sequential patterns. The experimental results show that, Concerning the sensitive regionsof space-time privacy inference attack, based on anonymous data association rules mining methodis proposed in this paper and the sequence mode, resulting in more threatening privacy risk.(3) firstproposed by dynamic analysis of the association rule privacy sensitive and sequence pattern, designoptimization of spatio-temporal K-anonymous privacy protection method on trusted servers, inorder to realize the basic idea of strict LBS location privacy protection. Finally, the results of this research have important practical value for the optimization ofspatio-temporal K-anonymous method to achieve enhanced protection of location privacy; and alsohave a certain theoretical reference value for researching privacy preserving data mining.