An Online LBS Privacy Protection Based On Dynamically Hiding Sensitive Sequential Rules Mined From Large Scale Of Anonymity Datasets

Recently, Location Based Services(LBS) has developed rapidly with the evolution and integration of mobile communication technology, positioning technology and geographic information technology. More mobile data with spatial-temporal attribute has been produced since LBS gets large-scale deployments. LBS application servers render pervasive services for different users by utilizing users’ request transaction data and extracting potential rules. However such analysis can also raises serious privacy concern. Based on a spatial-temporal correlation between large-scale datasets, an adversary can find sequence rules of anonymous query behavior, thus access users’ privacy. Prior privacy techniques based on spatial-temporal K-anonymity only meet the data quality and privacy requirement for successive updates, but ignore deep analysis based on large scale of spatial-temporal K-anonymity datasets. This raises the problem of inference attacks analyzing by large scale of spatial-temporal K-anonymity datasets, such as location prediction inference attacks.To overcome these challenges, we propose a spatial-temporal K-anonymity model by dynamic hiding sensitive sequence rules which achieve both the guaranteed location privacy of users and high data quality. The model first mines spatial-temporal sequence rules from anonymity datasets and analyzies inference attacks scenario offline, and then processes anonymous request and generates anonymity datasets in trusted anonymity server, and finally restarts mining spatial-temporal sequential rules from anonymity datasets and analyzing inference attacks scenario while the threshold is met. In this model, we design two different privacy algorithms for general sensitive sequence rules and sensitive sequence rules with sensitive areas(both are sensitive spatial-temporal sequential rules). The computational performance of suggested algorithms has been evaluated for its efficiency and validated by implementing them with a case study of automotive traffic monitoring datasets.