Research On Quality Of Service Prediction Algorithm Based On Differential Privacy

With the rapid development of big data,the recommender system has received extensive attention as an effective solution to the problem of information overload and passive service recommendation.As the research further develops,the application field of the recommender system is also expanding,extending from e-commerce,digital library,and other areas to Web services recommendation.In order to implement personalized Web service recommendation,it is necessary to collect and predict the situation in which users use Web services,that is,quality of service(Qo S).Most of the current work only focuses on how to improve the accuracy of Qo S prediction,but ignores the privacy protection of users.In fact,while the recommender system collects and mines personal characteristics and behavioral preferences to provide users with personalized recommendations,it also uses Qo S data to mine other sensitive user information,thereby exposing the user's personal privacy.In this paper,we first propose a collaborative Qo S prediction algorithm based on differential privacy to solve the problem of user privacy leakage.We propose two methods to disturb user data.The first one disturbs user data by applying differential privacy directly to user data.In order to increase the availability of disturbing data,the second method is to add noise satisfying differential privacy on aggregated user data.Then,in order to improve the accuracy of collaborative Web service Qo S prediction,we propose a shared collaborative service prediction algorithm based on differential privacy.This algorithm helps improve prediction accuracy by collecting data from multiple platforms of recommender system.However,each platform is not willing to share its own internal user data directly to the other party due to data privacy issues.In the process of sharing data,two methods of disturbing the shared data were proposed,and then a method of combining data owners and sharing data was proposed to perform secure prediction algorithms.In the experimental part,we evaluate the performance of the above two algorithms on a real web service data set.Experimental results show that the collaborative Web service Qo S prediction algorithm based on differential privacy still maintains good prediction accuracy under the guarantee of ensuring data privacy.The experimental results of the shared collaborative Web service Qo S prediction algorithm based on differential privacy prove that using the data of different platforms of recommender system to help the recommendation indeed improves the prediction accuracy,and the data security of different platforms is also protected.