Research On Technologies Of Users' Privacy Preserving In Location Based Services

Recent years have witnessed an increasing trend of global positioning system and mobile internet technology,people can easily enjoy a variety of location-based services(LBS)through smart mobile devices.Location data,which are used by LBS to provide various value-added services for users,reflects users'behavior habits,hobbies,etc.,while raising severe threat to users'privacy.Therefore,the privacy issues of users in LBS are nonnegligible.This paper mainly focuses on the research of users'privacy preserving techniques in LBS,including the privacy preserving issue of both users and the localization server in LBS location process,and the privacy attacks against users when requesting LBS.Specifically,the main work and innovations in this paper are as follows:1.This paper proposes a lightweight privacy-preserving scheme in Wi Fi fingerprint-based indoor localization(LWP~2).In Wi Fi fingerprint indoor localization algorithm,users'location data and fingerprints data of the localization server are all sensitive data that needs to be protected.Most related works ignore the data privacy of the localization server,and generates a large amount of overhead which is difficult to apply to terminal devices with limited computing resources.To this end,LWP~2formalizes the privacy-preserving indoor localization problem as minimizing the least-squared-error for an overdetermined linear formulation,and then design a lightweight solution in ciphertext space using the special structure of the overdetermined linear formulation.LWP~2,with considerable privacy performance,can successfully resist location privacy attacks and data privacy attacks without reducing the localization accuracy.2.This paper proposes Loc MIA,a more invasive attack system that allows adversaries to launch membership inference attacks against aggregated location data without reliance on any prior knowledge of the locations of victims.At present,the existing studies concerning membership inference attack mainly against genetic data,and fewer studies against location data suffer from the assumption that adversaries know the exact data of victims,which is impractical.Loc MIA is the first membership inference attack system to breach the membership privacy of individuals in LBS data aggregation.In particular,Loc MIA utilizes the impact of social relationships on mobility pattern to synthesize the victim's trajectory,and formalizes the membership inference problem as a binary classification problem by training the binary classifier with a supervised algorithm in machine learning.Finally,the privacy leaks caused by the proposed Loc MIA are nonnegligible,even in a weak adversarial knowledge setting.This provides certain reference significance for the design of privacy-preserving scheme in future LBS aggregation scenarios.