GNN4FR: A Lossless GNN-based Federated Recommendation Framework

TL;DR

This paper introduces GNN4FR, a novel federated learning framework for recommender systems that preserves user privacy while enabling full-graph training with high-order structural information, matching the performance of centralized methods.

Contribution

It is the first to propose a lossless federated GNN framework that constructs a global graph without leaking private user data, ensuring equivalent training to un-federated models.

Findings

Achieves full-graph training with privacy preservation.

Demonstrates equivalence to centralized GNN training.

Uses LightGCN as a case study.

Abstract

Graph neural networks (GNNs) have gained wide popularity in recommender systems due to their capability to capture higher-order structure information among the nodes of users and items. However, these methods need to collect personal interaction data between a user and the corresponding items and then model them in a central server, which would break the privacy laws such as GDPR. So far, no existing work can construct a global graph without leaking each user's private interaction data (i.e., his or her subgraph). In this paper, we are the first to design a novel lossless federated recommendation framework based on GNN, which achieves full-graph training with complete high-order structure information, enabling the training process to be equivalent to the corresponding un-federated counterpart. In addition, we use LightGCN to instantiate an example of our framework and show its equivalence.