Federated User Preference Modeling for Privacy-Preserving Cross-Domain Recommendation

TL;DR

A novel Federated User Preference Modeling (FUPM) framework is proposed to learn users' comprehensive preferences from both interaction data and additional data including review texts and potentially positive items, and the superiority of FUPM over SOTA baselines is validated.

Abstract

Cross-domain recommendation (CDR) aims to address the data-sparsity problem by transferring knowledge across domains. Existing CDR methods generally assume that the user-item interaction data is shareable between domains, which leads to privacy leakage. Recently, some privacy-preserving CDR (PPCDR) models have been proposed to solve this problem. However, they primarily transfer simple representations learned only from user-item interaction histories, overlooking other useful side information, leading to inaccurate user preferences. Additionally, they transfer differentially private user-item interaction matrices or embeddings across domains to protect privacy. However, these methods offer limited privacy protection, as attackers may exploit external information to infer the original data. To address these challenges, we propose a novel Federated User Preference Modeling (FUPM) framework. In FUPM, first, a novel comprehensive preference exploration module is proposed to learn users' comprehensive preferences from both interaction data and additional data including review texts and potentially positive items. Next, a private preference transfer module is designed to first learn differentially private local and global prototypes, and then privately transfer the global prototypes using a federated learning strategy. These prototypes are generalized representations of user groups, making it difficult for attackers to infer individual information. Extensive experiments on four CDR tasks conducted on the Amazon and Douban datasets validate the superiority of FUPM over SOTA baselines. Code is available at https://github.com/Lili1013/FUPM.

Figures (8)

• Figure 1: (a) Traditional CDR methods typically map and transfer user embeddings directly, without considering user privacy. (b) Existing PPCDR methods rely solely on user-item interaction histories to learn user embeddings and transfer differentially private embeddings across domains. (c) FUPM first utilizes review texts and potentially positive items to learn comprehensive user preferences and then privately transfers these preferences using differentially private prototypes within the FL framework.
• Figure 2: The overall framework of FUPM. It contains four modules: (a) Representation Learning Module aims to learn embeddings for user/item IDs and review texts. (b) Comprehensive Preference Exploration Module focuses on exploring comprehensive user preferences within each domain. It further divides into two components: (b).1 Contrastive feature alignment aims to align ID embeddings and review embeddings through CL. (b).2 Potential Interest Mining leverages potentially positive items to capture user's potential interests. (c) Private Preference Transfer Module aims to privately transfer user preferences across domains. (d) Prediction Module predicts user preferences.
• Figure 3: The performance of different interaction densities.
• Figure 4: The performance of different $\eta$ and $C$.
• Figure 5: The performance of different $\gamma$.