Location Privacy Protected Recommendation System In Mobile-cloud

As the core of location-based services(LBS),the LBS-oriented recommendation systems,which suggest the points-of-interest(POIs)to users by analyzing the distribution of the user's previous points-of-interest,have attracted great interest from both academia and industry.Despite the convenience brought by the LBS-oriented recommendation systems,most of current systems require users to expose their locations,which give rise to a big concerning of the location privacy issues.Meanwhile,as the defacto LBS infrastructure,the mobile-cloud computing paradigm introduces new opportunities and challenges to solve the privacy issues in LBS-oriented recommendation systems.To this end,we propose a novel location-privacy protected scheme for mobile-cloud based recommendation system.The main work of this thesis is as follows:1.The data protection of LBS-oriented recommendation system based on mobile cloud.In this thesis,the protection mechanism adopts different storage methods according to the different sensitivity of data.In order to protect information during the transmission from mobile terminals to cloud servers,we also provide a secure transmission mechanism with anonymous protection and fuzzy protection for geographical data and time.2.The solution of load balancing between mobile terminals and cloud servers.In this thesis,we take the mobile terminal defects such as storage capacity,computing performance,power into consideration as well as the network transmission pressure.In order to achieve load balancing,the mobile terminal and cloud server are assigned tasks of data storage and computing by the system intelligently.3.The algorithm based on historical check-in data for LBS-oriented recommendation system.In this thesis,we analyze the categorical correlation and geographical correlation by using kernel density estimation method to predict user behavior patterns,and then compute the score of unvisited POIs according to the user's behavior pattern for generating the recommendation result.4.The recommendation model and algorithm based on social relationship for LBS-oriented recommendation system.In this thesis,we present a new model combining the FOF model and the common concern of friends.There are two social influence here,one is the influence of similar users and the other is the influence of friends.As an important feature,the interaction between users will be used to compute the score of unvisited POIs as well as those two influence by adopting a new algorithm based on Hyperlink-Induced Topic Search.