Research On Privacy Protection Approaches In Location-Based Service

With the rapid development of mobile Internet and positioning technology,smart mobile terminals gain ground.Users can easily obtain various location-based services(LBS)based on location attributes.In Chinese market,LBS has tens of billions of yuan application market,such as Sina Weibo,Meituan and other applications.The service system of LBS applications is based on the user's location attribute.When the user uses LBS application,his LBS query is composed of location information and content information.LBS query not only directly contains the user's whereabouts,but also implies the user's daily behavior patterns,such as home address,interest preferences and physical condition.Therefore,direct disclosure of the LBS query to an untrusted third party(service provider)may open the door to cause misuse of privacy information,and pose a serious threat to the user privacy.In order to solve the above problem,we firstly analyze how the LBS query is handled and used by the adversary,to guide how to protect the privacy of the LBS query as a whole.The LBS query is mainly attacked by the following two adversaries: 1)Online adversary: this kind of adversary own some knowledge,such as real-time traffic congestion,map information,weather conditions and other online information.These knowledge is taken advantage of to attack the LBS query protected by privacy protection methods.The adversary is hopeful that he can recognize the user's real LBS query.2)Offline adversary: to be able to discover the user's behavioral pattern,the user's historical query data is conducted with an offline fashion by the offline adversary.The adversary is hopeful that he could use the user's behavioral pattern to identify the user's real LBS query which is protected by privacy protection method.Based on different adversary features,this thesis carries out relevant research work on the framework of privacy protection and the involved methods in location-based service.The main research work of this thesis are organized as follows:1)In order to improve the expectation of the privacy protection effect in face of different types of adversary,this thesis uses the ?-anonymity theory to construct a privacy protection framework.The framework can flexibly switch the different privacy protection method according to the different category of the LBS query.Specifically,the framework is divided into four layers from the bottom up: positioning technology layer,mobile client,communication network layer and LBS provider.The positioning technology layer is responsible for providing real-time location for the mobile client;the privacy protection middleware is the core of the entire framework,deployed on the mobile client to implement different privacy protection methods to combat online or offline adversary;the communication network layer provides communication services between the mobile client and the LBS server;the LBS provider is responsible for processing LBS query.2)In order to improve the online protection effect for the LBS query,an online privacy protection method is proposed in this thesis.This method is based on the relevance of location information and content information in the LBS query.Specifically,we use existing privacy protection methods to generate dummy LBS query which content information is empty,and take advantage of the relationship between location information and content information to generate the content candidate set.Then these candidates are screened for semantic relevance,and finally combined into dummy LBS query.When the user makes a continuous query,the continuous dummy LBS query is checked for reachability and semantic relevance.After that the online protection effect trends to the ?-anonymity.Experiments based on the real data set GeoLife show that this proposed method has a better privacy protection effect than the comparison method against online adversary.3)In order to improve the offline protection effect for the LBS query,an offline privacy protection method is proposed in this thesis.This method is based on the user's behavior patterns contained in long-term(more than 30 days)LBS query data.Specifically,we use a hierarchical clustering algorithm based on the distance of the road network for the user's historical query data,to discovered the user's frequent areas(the areas where users frequently initiate queries,such as homes,companies,etc.).For the user's single LBS query,the dummy LBS query is generated considering whether the user is located in a frequent area;for the user's continuous LBS query,the consecutive dummy LBS query is performed with checking the time reachability.