Research On Privacy Protection In Mobile Sensing Recommender System

With the development of mobile Internet,users' data generated by mobile devices is growing at an unprecedented rate.In order to obtain the valuable information from severely overloaded resources and meet the personalized needs for users,the emergence of mobile-aware recommender system provides a new way for users to obtain information.The mobile-aware recommender system is usually built on massive user data.By analyzing user's historical behavior,the recommender system can obtain users' preference and realize the recommendation for users.However,current recommender systems mainly focused on high-quality recommendations while underestimating privacy issues,which can lead to problems of privacy.Such problems are more prominent when service providers,who have limited computational and storage resources,leverage on cloud platforms to fit in with the tremendous number of service requirements and users.This paper studies the key technologies of data security and privacy issues in mobile sensing recommender systems,including: privacy-preserving framework for location-aware mobile recommender system,adaptive privacy-preserving mechanism for continuous locationbased services;trust-based privacy-preserving framework for friend recommendation in mobile social network.The main research achievements can be summarized as follows:(1)To solve the data security and privacy issues in location-aware recommender systems,we propose a novel framework for protecting user privacy information,including locations and recommendation results,within a cloud environment.Through this framework,all historical ratings are stored and calculated in ciphertext,allowing us to securely compute the similarities of venues through Paillier encryption,and predict the recommendation results based on Paillier,commutative,and comparable encryption.However,since the Paillier semi-homomorphic encryption is introduced,the above scheme has a lower efficiency.To address the problem,a lightweight privacy-preserving framework for location-aware recommender system is proposed.Through the framework,service providers upload the historical rating data,which are processed with random function to cloud platform and compute the similarities of venues securely.Then,users encrypt the interesting area with comparable encryption and predict the recommendation results with a carefully designed secure protocol.Consequently,the lightweight privacy-preserving framework is able to protect the privacy of users and the property of service providers.Compared with the first work based on homomorphic encryption methods,the lightweight scheme is more efficient and has a better user experience.(2)The differential privacy mechanism only works well in the case of snapshot locationbased services(LBSs),which would not apply to the case of continuous LBSs due to the quick loss of privacy caused by the correlation between locations in the trace.To solve the problem,we propose a novel mechanism to protect user's location privacy in continuous LBSs.In our mechanism,a R-tree is introduced to realize the reusable of generated sanitized locations,which achieves the differential privacy with less consumption of privacy budget.Finally,empirical results over a real-world dataset demonstrate that with the same utility,our mechanism consumes less privacy budget to obfuscate the same trace.(3)To protect the privacy information in friend recommendation in online social network(OSN),we propose a novel decentralized framework which utilizes OSN users' social attributes and trust relationships to achieve the friend recommendation in a privacy-preserving manner.In our mechanism,we adopt a light-weight privacy-preserving protocol to aggregate the utilities of multi-hop trust chains and compute the recommender results securely.We also analyze the efficiency of our mechanism in theory.Finally,we conduct an experiment to evaluate our mechanism over a real-world dataset and empirical results demonstrate that our mechanism can effectively and efficiently recommend friends in a privacy-preserving way.