Research On Spatio-temporal Features Based Location Privacy Inference In Social Networks

With the social networks flourish,the public generally used social apps to record and share their daily lives.As same time,social applications also provide a variety of powerful social services.The check-in service is one of the mainstream social networking applications.In the check-in service,the user uses the mobile device with positioning function to send the server their geographical location for checking in.The check-in data has the features of spatiotemporal and reflects the user's behavior patterns.At the same time,user's behavior pattern can be tapped based on these check-in data to provide a new dimension for location deduction.By analyzing the existing location inference method,it is found that the new location can not be deduced due to the sparseness of trajectory data and implicit location access in the current location deduction.The accuracy of location visit is low,but some methods can deduce the user's implicit location.However,there is a problem that requires additional preparation knowledge or computational overhead.After analyzing the existing methods of mining user behavior patterns,it is found that the mining results of current Apriori-like methods lose the spatiotemporal correlation between sign-in points.However,the time series Frequent pattern mining time overhead and unacceptable.Aiming at above problems,this paper proposes a check-in sequence pattern based hiddenlocation inference(marked PBHLI).Firstly,it elaborates on the related knowledge and problem definitions involved in the user registration mode and implicit position deduction,including attendance location algorithm and continuous subsequence algorithm in sub trajectory,which laid the foundation for the implementation of the research on spatio-temporal features based location privacy inference in social networks.Secondly,in the PBHLI algorithm,the check-in sequence pattern is generated by using the historical trajectory data and its check-in probability is calculated.Based on the check-in sequence pattern probability,propose a check-in sequence extend algorithm,the access trajectory data that may contain hidden location is extended by this algorithm.The number of candidate sets is increased by the check-in sequence extend algorithm,which reduces the sparseness of trajectory data and improves the precision of hidden location access probability calculation.Finally,a hidden location deduction algorithm based on user check-in pattern using Bayesian model is designed.The algorithm can effectively infer the hidden location.Through the experiment of real data sets compared with existing methods,it is verified that the proposed algorithm based on the check-in sequence pattern implies that the user's hidden location and non-hidden location correct rate,and has a good time overhead.