PRSI: Privacy-Preserving Recommendation Model Based on Vector Splitting and Interactive Protocols

TL;DR

This work addresses privacy leakage in recommendation systems when each client holds data for a single user. It introduces PRSI, a privacy-preserving framework based on vector splitting and interactive protocols, consisting of a preprocessing module and two phases for collecting interactions and delivering recommendations. The main contributions are the vector-splitting technique with fake interactions, a secure data-collection protocol that hides IDs and IPs, and a result-distribution protocol that preserves privacy while enabling accurate training and delivery; experiments on Yelp2018 demonstrate security against partial data leakage, acceptable transmission accuracy, and scalable communication costs. PRSI offers a practical privacy layer that can be flexibly integrated with existing recommendation models and potentially combined with federated learning for enhanced real-world applicability.

Abstract

With the development of the internet, recommending interesting products to users has become a highly valuable research topic for businesses. Recommendation systems play a crucial role in addressing this issue. To prevent the leakage of each user's (client's) private data, Federated Recommendation Systems (FedRec) have been proposed and widely used. However, extensive research has shown that FedRec suffers from security issues such as data privacy leakage, and it is challenging to train effective models with FedRec when each client only holds interaction information for a single user. To address these two problems, this paper proposes a new privacy-preserving recommendation system (PRSI), which includes a preprocessing module and two main phases. The preprocessing module employs split vectors and fake interaction items to protect clients' interaction information and recommendation results. The two main phases are: (1) the collection of interaction information and (2) the sending of recommendation results. In the interaction information collection phase, each client uses the preprocessing module and random communication methods (according to the designed interactive protocol) to protect their ID information and IP addresses. In the recommendation results sending phase, the central server uses the preprocessing module and triplets to distribute recommendation results to each client under secure conditions, following the designed interactive protocol. Finally, we conducted multiple sets of experiments to verify the security, accuracy, and communication cost of the proposed method.

Figures (7)

• Figure 1: Basic steps of Vector Splitting.
• Figure 2: The Jaccard similarity between the speculated vector $\bm{u}{spe}$, calculated by client $i$ using split vectors, and the interaction vector $\bm{u}{i,2}^{**}$.
• Figure 3: The impact of the ratio $c$ of fake interaction items to real interaction items on the Jaccard similarity.
• Figure 4: The relationship between the number of digits in the virtual IDs and the repetition rate.
• Figure 5: The impact of the attenuation factor $\alpha$ on communication cost during the collecting interaction vectors phase and the sending recommendation results phase.