A Privacy-Preserving Subgraph-Level Federated Graph Neural Network via Differential Privacy

Currently, the federated graph neural network (GNN) has attracted a lot of attention due to its wide applications in reality without violating privacy regulations. Among all privacy-preserving technologies, differential (DP) is most promising one effectiveness and light computational overhead. However, DP-based GNN not been well investigated, especially sub-graph-level setting, such as scenario recommendation system. The biggest challenge how guarantee solve non independent identically distributed (non-IID) data simultaneously. In this paper, we propose DP-FedRec, fill gap. Private Set Intersection (PSI) leveraged extend local for each client, thus non-IID problem. Most importantly, DP applied only on weights but also edges intersection from PSI fully protect clients. evaluation demonstrates DP-FedRec achieves better performance with extension introduces little computations