Research On Privacy-preserving Spatio-textual Similarity Joins

With the GPS popularization in mobile devices,the number of applications based on location-based services increases rapidly.Promoting the social recommendation and custom segmentation,spatio-textual similarity joins have been widely applied in mobile Internet and mobile social media.To reduce the costs of data processing and management,data owners are motivated to process and manage their data on cloud platform.However,it may arise privacy concerns in the data outsourcing for spatio-textual similarity joins.To address this problem,in this thesis the privacy requirements of spatio-textual similarity joins in outsourced scenario are deeply analyzed,and the privacy-preserving spatio-textual similarity joins in cloud environments are studied.Firstly,a secure spatio-textual similarity joins method based on an inner-product preserving encryption is proposed.In this method,according to calculation features of spatio-textual similarity joins,the spatio-textual data are transformed into vectors and encrypted using the inner-product preserving encryption algorithm.Then,exploiting its inner-product preserving manner,the secure computation of spatio-textual similarities can be achieved.At last,with the computed similarities,the data objects are joined into pairs with a higher similarity to achieve the privacy-preserving spatio-textual similarity joins.To further improve the efficiency of spatio-textual similarity joins,a secure spatio-textual similarity joins method based on locally sensitive hashing(LSH)is presented.In this method,a secure index structure based on the encryption of a hybrid LSH is first constructed.Then,with the similar LSH mechanism,the hybrid LSH values for the request data of spatio-textual similarity joins are computed and encrypted.At last,searching over the secure index with the encrypted hash values,the candidate set of spatio-textual data can be obtained,and the secure inner-product computation can be done to calculate the spatio-textual similarities over the candidate set and the request data set,which are encrypted by the inner-product preserving encryption,so that pairs of the data objects with a higher similarity can be joined.Thus,the privacy-preserving spatio-textual similarity joins are achieved efficiently and securely.Finally,the two proposed methods are analyzed theoretically from two aspects of security and complexity.Besides,the experimental results on real data sets show that the proposed LSH-based method can effectively improve the efficiency of the privacy-preserving spatio-textual similarity joins.