A Trust-Delivery Collaborative Recommendation Method Based on Local Differential Privacy
Wanqing Wu, Xin Mi Yang, Ruohe Lei · 2023
The collaborative filtering recommendation system provides personalized recommendation service for users by collecting a large amount of user information, but there is a risk of personal privacy leakage in this process. Therefore, it is critical to address the trade-off between recommendation performance and privacy. To solve this problem, this paper proposes a recommendation method based on trust values and local differential privacy, which is specified as follows. First, this paper designs a recommendation model based on the trust delivery method, which uses trust values to measure the similarity between users, and a neighbor set of the target user is selected from the trust value matrix, so as to perform rating prediction and provide a recommendation list. Then, it proposes a privacy recommendation model that normalizes the user’s rating data to reduce global sensitivity and uses a double perturbation method to achieve privacy protection for both user’s rating behavior and rating data. Finally, the proposed algorithm is proved to satisfy ε-differential privacy by security analysis, and effectiveness of this algorithm is evaluated on the real data set. Experimental results show that the proposed method improves recommendation performance and data availability while ensuring privacy in comparison with previous privacy protection schemes for recommendations.