Research On Location Privacy Protection In Location Social Network Based Personalized Recommendation

Recently,due to the popularity of mobile devices and the rapid development of wireless positioning technologies,location-based recommendation services have attracted increasing attentions.Based on users' submitted location data,the recommendation server can provide personalized recommendations for them,which brings significantly convenience and contributes to improving life quality of the users.However,location data implies much inferred sensitive information of the users.The collection of those location data can cause the users' sensitive information,such as daily activities,health condition and social status,be mined and exposed.Therefore,how to protect the users' location privacy is a problem to be solved,and heavily hinders the development of location-based recommendation.Most existing schemes of location privacy protections in recommendation systems are proposed under the assumption that users trust their recommender that will manage their users' private data properly.And their aims are to protect the privacy in the process of data released to the third party.However,in reality,there is no trusted recommendation server,and the server may analysis the users' sensitive information that they do not want others to discover,thus resulting in further disclosure of users' personal information.In this thesis,we have a deep research on the location privacy protection in social network based personalized recommendation systems.We propose two privacy protection schemes for discrete location points and trajectories under the untrusted recommender systems.The main contents are listed as follows:1.We have studied the background,current situation and development trend of the location privacy in the personalized recommendation systems,and have grasped the hot issues and relevant academic progress in the related fields,and thus determined the subject of our research.We study the existing schemes and give a brief introduction of the common mechanisms,techniques and schemes.And thus we understand the major challenges and mainstream protection technologies in the research of location privacy protection of recommendation.In addition,we study on the design objective,system architecture and threat model of the location-based services.2.We propose a mobile devices-based location privacy algorithm to generate dummy locations relying on the location semantic similarity and spatial cloaking,in order to protect the privacy of location points in the personalized recommendation system.Our schemes solve the shortcomings of the generation of unreasonable dummy locations in the existing researches.We consider the spatiotemporal attributes of location data,and the semantic concept distance(SCD)method is employed to measure the similarity between the semantics of location data.In addition,to ensure that the distribution of the generated dummy locations and the true location does not reveal the exact range of the user's position,we utility the spatial cloaking technology to select dummy locations from the candidate dummy locations shared with similar semantic.3.In order to solve the problem of trajectory privacy protection in personalized recommendation,we propose a fake trajectory generation algorithm based on the Hidden Markov Model(HMM)and semantic clustering.Our solution mainly solves the shortcomings of the existence of trusted third parties in the existing fake trajectory privacy scheme and the disclosure of other users' trajectory information.The fake but similar trajectory we generated is not the real trajectory of the other users in the database,but the position sequence composed of different location points.And due to the limitations of client storage and computing power,we use the cloud database to generate fake trajectory,for its huge data storage and powerful computing power,and then send the generated fake trajectory to the users to obtain the personalized recommender.4.Finally,through theoretical analysis and experimental simulation,we demonstrate the feasibility of our two proposed schemes on the real-world dataset collected by Foursquare.And our schemes can reduce the amount of users' sensitive information that the recommender server obtains from the users' submitted location sets,and meet the users' desire to protect their location privacy,while obtaining personalized recommender.