Differential private collaborative Web services QoS prediction

摘要: Collaborative Web services QoS prediction has proved to be an important tool estimate accurately personalized experienced by individual users, which is beneficial for a variety of operations in the service ecosystem, such as selection, composition and recommendation. While number achievements have been attained on study improving accuracy collaborative prediction, little work done protecting user privacy this process. In paper, we propose privacy-preserving framework can protect private data users while retaining ability generating accurate prediction. We introduce differential privacy, rigorous provable model, into process first present DPS, method that disguises user’s observed values applying directly. show how integrate DPS with two representative approaches. To improve utility disguised data, DPA, another disguising aggregates before adding noise achieve privacy. evaluate proposed methods conducting extensive experiments real world dataset. Experimental results our approach feasible practice.