[Paper Review] Secure Social Recommendation based on Secret Sharing

Chaochao Chen, Liang Li+4 more|arXiv (Cornell University)|Feb 6, 2020

Recommender Systems and Techniques37 references14 citations

TL;DR

This paper proposes SeSoRec, a secure multi-party computation framework that enables collaborative social recommendation between a rating platform and a social platform without exposing raw user data. By introducing a novel Secret Sharing-based Matrix Multiplication (SSMM) protocol, SeSoRec achieves linear computation and communication complexity while preserving privacy, outperforming existing methods in accuracy and efficiency on real-world datasets.

ABSTRACT

Nowadays, privacy preserving machine learning has been drawing much attention in both industry and academy. Meanwhile, recommender systems have been extensively adopted by many commercial platforms (e.g. Amazon) and they are mainly built based on user-item interactions. Besides, social platforms (e.g. Facebook) have rich resources of user social information. It is well known that social information, which is rich on social platforms such as Facebook, are useful to recommender systems. It is anticipated to combine the social information with the user-item ratings to improve the overall recommendation performance. Most existing recommendation models are built based on the assumptions that the social information are available. However, different platforms are usually reluctant to (or cannot) share their data due to certain concerns. In this paper, we first propose a SEcure SOcial RECommendation (SeSoRec) framework which can (1) collaboratively mine knowledge from social platform to improve the recommendation performance of the rating platform, and (2) securely keep the raw data of both platforms. We then propose a Secret Sharing based Matrix Multiplication (SSMM) protocol to optimize SeSoRec and prove its correctness and security theoretically. By applying minibatch gradient descent, SeSoRec has linear time complexities in terms of both computation and communication. The comprehensive experimental results on three real-world datasets demonstrate the effectiveness of our proposed SeSoRec and SSMM.

Motivation & Objective

• To address the challenge of improving recommendation performance using social data from external platforms while preserving data privacy.
• To formalize secure social recommendation as a secure multi-party computation (MPC) problem.
• To design a practical, scalable, and privacy-preserving framework that enables collaboration between platforms with sensitive data.
• To develop a secret sharing-based matrix multiplication protocol (SSMM) that eliminates reliance on trusted third parties.
• To empirically validate the framework's effectiveness and efficiency on real-world datasets.

Proposed method

• Formalize secure social recommendation as a multi-party computation problem where two platforms (rating and social) jointly learn user preferences without exposing raw data.
• Propose SeSoRec, a framework that combines user-item ratings from platform 𝒜 with social relationships from platform 𝒷 using collaborative learning.
• Design a novel Secret Sharing-based Matrix Multiplication (SSMM) protocol to securely compute matrix products in the MPC setting, replacing trusted initializers.
• Use secret sharing to distribute matrix elements among parties, enabling secure inner product computation without revealing inputs.
• Apply minibatch gradient descent to ensure linear time complexity in both computation and communication for scalability.
• Prove the correctness and security of SSMM under semi-honest threat model, ensuring no information leakage beyond the output.

Experimental results

Research questions

• RQ1 Can a secure framework be designed to enable collaboration between a rating platform and a social platform without exposing raw user data?
• RQ2 How can matrix multiplication—critical for collaborative filtering—be securely computed in a multi-party setting without a trusted third party?
• RQ3 Does the proposed SSMM protocol achieve better efficiency than existing trusted-initializer-based secure matrix multiplication protocols?
• RQ4 To what extent does incorporating social information improve recommendation accuracy, especially in sparse data regimes?
• RQ5 How does the social regularizer parameter γ affect the balance between rating and social information in the final model?

Key findings

• SeSoRec achieves recommendation accuracy nearly identical to the unsecured Soreg model, with RMSE differences of less than 0.0003 across all datasets, validating the correctness of the SSMM protocol.
• On the Epinions dataset (0.48% rating density), SeSoRec improves MF by 7.60% in RMSE, demonstrating significant gains in sparse settings.
• SSMM reduces running time by approximately 2.4× compared to TISMM, with 0.0025s for dimension 100, 0.3246s for 1000, and 40.744s for 10000, showing strong scalability.
• The performance of SeSoRec is highly sensitive to the social regularizer parameter γ, with optimal performance achieved at γ = 10^0 on the FilmTrust dataset.
• SeSoRec maintains linear computation and communication complexity, enabling efficient deployment on large-scale datasets.
• The SSMM protocol eliminates the need for a trusted initializer, enhancing practicality and security in real-world deployments.