Secure Federated Graph-Filtering for Recommender Systems

2025-01-28

AI Summary

Centralized aggregation of user-item interaction data in graph-based recommenders—e.g., via normalized adjacency matrices or low-pass graph filters—poses severe privacy and security risks. Method: We propose two decentralized frameworks for secure distributed computation of critical graph components. Our approach introduces a novel privatization paradigm integrating lightweight multi-party computation (MPC) with distributed singular value decomposition (SVD), and the first low-rank approximation mechanism enabling controllable trade-offs between communication overhead and prediction accuracy. Results: Evaluated on multiple benchmark datasets, our method matches centralized state-of-the-art (SOTA) accuracy while reducing communication costs by over 40%. It strictly ensures that raw user data remains on-device (no data leaves the local domain) and that model inputs are provably irreversible—achieving strong formal privacy guarantees without compromising practical utility.

Technology Category

• Machine Learning: Privacy
• Search and Optimization: Distributed Search
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Recommender systems often rely on graph-based filters, such as normalized item-item adjacency matrices and low-pass filters. While effective, the centralized computation of these components raises concerns about privacy, security, and the ethical use of user data. This work proposes two decentralized frameworks for securely computing these critical graph components without centralizing sensitive information. The first approach leverages lightweight Multi-Party Computation and distributed singular vector computations to privately compute key graph filters. The second extends this framework by incorporating low-rank approximations, enabling a trade-off between communication efficiency and predictive performance. Empirical evaluations on benchmark datasets demonstrate that the proposed methods achieve comparable accuracy to centralized state-of-the-art systems while ensuring data confidentiality and maintaining low communication costs. Our results highlight the potential for privacy-preserving decentralized architectures to bridge the gap between utility and user data protection in modern recommender systems.

Problem

• Recommendation Systems
• Data Privacy
• Data Security

Innovation

• Secure Federated Graph Filtering
• Privacy Protection
• Efficient Personalization