Research on Matrix Factorization Recommendation Algorithm Based on Local Differential Privacy

Mobile Edge Computing (MEC) has gained significant attention in enhancing the efficiency of Recommendation systems. However, the trustworthiness of servers poses a challenge as they can potentially compromise user privacy. To address this issue, we propose a framework for matrix factorization-based recommendation using Local Differential Privacy (LDP). Initially, user data is perturbed using Piecewise Mechanism (a kind of LDP algorithm) and published to an edge server. The edge server performs basic computations on the perturbed data, while the cloud server employs matrix factorization to compute latent factors for users and items, which are then sent back to the edge server. Finally, the edge server computes similarity values and generates personalized recommendations for users. Through extensive simulations, our algorithm ensures recommendation accuracy while preserving user privacy. By comparing with the generalized differential privacy mechanism, the Piecewise Mechanism used in this paper has a better recommendation effect, thereby demonstrating its practical utility.