Research On Differential Privacy Method For Location Based Service

With the rapid development of smart devices,it has become a common phenomenon to optimize resources and provide better services by collecting,mining and analyzing user location information.Existing location-based services provide reliable decision support for national construction,urban management,public services,etc.However,in front of un-trusted third-party service providers,collecting users' spatial location information may cause users' privacy leakage,because combining with some background knowledge,analyzing and processing location data directly can easily infer users' personal privacy.Then users are unwilling to share and use such services if the user's personal location information cannot be guaranteed,which will make it difficult to develop the location based services,and lead to economic losses of society.The "Proposal" to "Outline" of China's "14th Five-Year Plan" both proposed new requirements for network security under the digital transformation.The 13 th Five-Year Plan emphasizes the importance of the privacy security of spatial location data of "internet plus Big Data/Cloud Computing".Therefore,privacy protection of location data is the key to ensure the healthy and sustainable development of the location service industry.In the aspect of location-based data privacy protection,the current algorithms mainly include privacy policy,encryption mechanism and obfuscation mechanism,among which obfuscation mechanism includes anonymity,obfuscation(differential privacy),dummy locations and so on.Differential privacy(DP)was proposed by Dwork(2006)and has becomes a hot research field of data privacy protection.It is based on a meaningful and mathematically rigorous definition of a probabilistic privacy model and independently from any side information that an attacker might possess.The current differential privacy algorithm based on location data solves the privacy protection problem in some practical applications.However,due to the characteristics of location data and the different privacy level in different applications,the current differential privacy algorithm cannot be entirely adopted to application.The existing research mainly has the following problems:(1)Due to the influence of positioning technology,application scenarios,environment and other factors,there are inevitable positioning errors in location data.But there is no analysis of the impact of errors on differential privacy protection.(2)The trade-off between privacy and utility depends closely on the application type,and the evaluation of privacy and utility should be adjusted according to the actual application.At present,the research of differential privacy algorithm is not enough.In view of the current problems,the main research of this paper is summarized as follows:(1)The existing differential privacy methods are based on the original data is accurate and without considers errors.However,due to the influence of positioning accuracy,the errors in the location data make the privacy and service utility under differential privacy methods deviate from the theoretical expectations.In this paper,we focus on the influence of location error on the mechanism of differential privacy;we also proposed a mix mechanism,which realizes the privacy analysis of location error on privacy protection and provides theoretical support for the practical application of location data.(2)Based on the travel time prediction of typical applications,a time prediction differential privacy protection algorithm without collecting the original location data is proposed.Because the third party collects and analyzes the user's location information,it inevitably causes the privacy leakage of the user's location information.In order to ensure the balance between privacy security and service utility,several utility evaluation methods are proposed.On this basis,the evaluation method of resisting adversary re-recognition is proposed,which realizes the privacy protection of users in typical applications of time prediction and provides reference for privacy configuration parameters for current applications.(3)For the typical application of location tracking,monitoring users' location will expose users to attacks,which will lead to privacy leakage.Based on location tracking under the outbreak of COVID-19,firstly,we summarizes the current location tracking technologies and privacy problems faced by various countries and governments in the current outbreak of COVID-19;Then we proposes privacy protection methods and evaluation means for multiple location publishing based on Geo-indistinguishable differential privacy to realizing the privacy protection of users when publishing their own locations.At the same time,the randomized response algorithm of differential privacy and the Geo-indiscernibility algorithm are adopted to evaluate the privacy and utility of location tracking social network publishing.The proposed method realizes the privacy protection of location social network among users under epidemic situation,and provides theoretical support and privacy configuration parameter reference for privacy protection in epidemic tracking.In this paper,aiming at the privacy protection problems in location-oriented data applications and the shortcomings of current privacy protection algorithms,a differential privacy protection algorithm considering errors is proposed;For two typical applications of location data,the corresponding differential privacy protection methods are presented,and the corresponding availability evaluation methods are designed,which achieve a good balance between the privacy and availability of location data applications.Solving the problem of application-oriented privacy protection of differential privacy is not only the extension of differential privacy theory,but also has important practical significance and practical value for the application of differential privacy.