Elevating Privacy in Recommendation Systems with Hybrid Noise in Local Differential Privacy

This article addresses privacy concerns in recommendation systems, crucial for content discovery, by exploring privacy-preserving techniques such as Local Differential Privacy (LDP) to balance user privacy and suggestion accuracy. Focusing on matrix factorization, collaborative filtering, and multi-domain recommendation systems, the study analyzes trade-offs between suggestion accuracy and user privacy. The proposed Hybrid Noise method, incorporating Laplace and Bounded Laplace noise, enhances privacy without compromising recommendation accuracy. Empirical evaluations reveal improved outcomes, evidenced by promising F1-scores, MAE, and RMSE results. The research contributes insights into the current state of privacy in Recommendation Systems and introduces Hybrid Noise as a comprehensive approach to balancing user privacy and recommendation quality, showcasing significant enhancements in both areas. These findings pave the way for future research at the intersection of privacy preservation and customization in dynamic digital landscapes.