Federated Heterogeneous Graph Neural Network for Privacy-preserving Recommendation

TL;DR

This work addresses privacy-preserving recommendations over heterogeneous information networks in a federated setting by partitioning the HIN into private client data and shared server data. It introduces FedHGNN, which combines a semantic-preserving two-stage publishing mechanism (EM-based selection of relevant shared HINs and degree-preserving randomization within semantic-guided items) with a heterogeneous graph neural network that performs meta-path–guided node and semantic aggregations. The authors formalize epsilon-semantic privacy and epsilon-semantic guided interaction privacy, provide privacy guarantees for the publishing process, and demonstrate that FedHGNN substantially outperforms FedRec baselines (up to 34% HR@10 and 42% NDCG@10) and is competitive with centralized HIN/GNN methods. The approach offers a practical pathway to privacy-conscious, semantically rich recommendations in distributed data environments, with robust empirical validation across four real-world datasets.

Abstract

The heterogeneous information network (HIN), which contains rich semantics depicted by meta-paths, has emerged as a potent tool for mitigating data sparsity in recommender systems. Existing HIN-based recommender systems operate under the assumption of centralized storage and model training. However, real-world data is often distributed due to privacy concerns, leading to the semantic broken issue within HINs and consequent failures in centralized HIN-based recommendations. In this paper, we suggest the HIN is partitioned into private HINs stored on the client side and shared HINs on the server. Following this setting, we propose a federated heterogeneous graph neural network (FedHGNN) based framework, which facilitates collaborative training of a recommendation model using distributed HINs while protecting user privacy. Specifically, we first formalize the privacy definition for HIN-based federated recommendation (FedRec) in the light of differential privacy, with the goal of protecting user-item interactions within private HIN as well as users' high-order patterns from shared HINs. To recover the broken meta-path based semantics and ensure proposed privacy measures, we elaborately design a semantic-preserving user interactions publishing method, which locally perturbs user's high-order patterns and related user-item interactions for publishing. Subsequently, we introduce an HGNN model for recommendation, which conducts node- and semantic-level aggregations to capture recovered semantics. Extensive experiments on four datasets demonstrate that our model outperforms existing methods by a substantial margin (up to 34% in HR@10 and 42% in NDCG@10) under a reasonable privacy budget.

Figures (6)

• Figure 1: Comparison of a HIN in the centralized setting and federated setting
• Figure 2: The overall framework of FedHGNN
• Figure 3: The two-stage perturbation mechanism for user-item interaction publishing
• Figure 4: Effects of different number $n$ of shared HINs
• Figure 5: Effects of different privacy budget $\epsilon_1$ and $\epsilon_2$

Theorems & Definitions (9)

• Definition 2.1
• Definition 2.2
• Definition 2.3
• Definition 2.4
• Definition 2.5
• Definition 2.6
• Definition 2.7
• Theorem 3.1
• Theorem 3.2