GDTM: Gaussian Differential Trust Mechanism for optimal recommender system

As recommender systems have become increasingly popular in providing users with personalized recommendations, researchers have implemented protective measures to safeguard users’ privacy. However, the implementation of such mechanisms is extremely difficult to ensure both recommendation accuracy and privacy protection. In this paper, we propose a novel protective mechanism that addresses this challenge. Our approach introduces the concept of differential trust, which integrates matrix factorization and the combination theorem of differential privacy. We then propose the Gaussian Differential Trust Mechanism, which protects users’ historical ratings while maintaining recommendation accuracy to a certain extent. The rationality of our proposed mechanism is verified by theoretical explanation and experimental evaluation. The experiment results demonstrate that our method effectively balances the competing goals of recommendation accuracy and privacy preservation, providing a solution to the challenges faced by recommender systems.