Differential Privacy Protection Scheme For Location-based Services

With the rapid development of mobile Internet,location-based services(LBSs)have also developed rapidly.LBSs require that the user sends his location to the service party,and the service party returns specific information according to the user's query.However,usually the service party is untrusted,and once the user's location information is illegally collected and shared,sensitive information such as the user's home address and work address may be exposed.Therefore,the protection of the user's location privacy needs to be resolved.Differential privacy is a popular method for location privacy protection method,geo-indistinguishability based on differential privacy is a strict concept of location privacy and widely used in the location privacy protection.However,the location protection scheme based on geo-indistinguishability can't effectively resist Bayesian attack when users are in isolated locations.In order to resist Bayesian attack,the personalized location privacy protection scheme sets geoindistinguishability parameters according to user's location.However this makes the user's location and geo-indistinguishability parameters have a certain correlation,the adversary attacks by uses the geo-indistinguishability parameter as background knowledge can obtain a more accurate set of the user's possible locations,which we call the geo-indistinguishability parameter attack.The main contributions of this thesis are as follows:(1)In this thesis,the problem of geo-indistinguishability parameter attack is first proposed in the context of personalized location privacy protection,to address this problem propose a PDBB mechanism that can effectively resist geo-indistinguishability parameter attack and Bayesian attacks.This thesis gives more losser security conditions,expands the protection region search range,resists geo-indistinguishability parameter attacks by adding noise to the privacy budget allocated to the protection region,and uses the redundant privacy budget generated by noise to enlarge the release probability to achieve the purpose of improvingutility.(2)Since the location protection scheme based on geoindistinguishability can't effectively resist Bayesian attack when users are in isolated locations,this thesis proposes a KDP mechanism combining geo-indistinguishability and k-anonymity,KDP mechanism can effectively resist Bayesian attack in isolated locations by hide the obscucated location in k-1 dummy location.