FedRGL: Robust Federated Graph Learning under Label Noise

Federated Graph Learning (FGL) is a distributed machine learning paradigm based on graph neural networks, enabling secure and collaborative modeling of local graph data among clients. However, label noise in graph data can degrade the generalization performance of the global model. Existing federated label noise learning methods, primarily focused on computer vision tasks, often yield suboptimal results when directly applied to FGL. To address this issue, we propose a robust federated graph learning method with label noise, termed FedRGL. Specifically, FedRGL leverages the globally aggregated model and local subgraph structural information to implement a dual-perspective consistency noise-node filtering mechanism under class-aware dynamic thresholds. The resulting class-aware dual-consistency filtering (CADF) can also serve as a plug-and-play module, enhancing noise robustness across various subgraph federated learning frameworks. To better exploit the supervisory information from filtered noisy nodes, we employ the natural augmentation techniques from graph contrastive learning to assign high-confidence pseudo-labels to the noise nodes. Additionally, we measure model quality via the average predictive entropy of unlabeled nodes, enabling adaptive robust aggregation on the server side. Extensive experiments on real-world graph datasets show that FedRGL consistently outperforms existing methods under different noise rates, noise types, and client scales, achieving on average 5--8\% higher accuracy and up to 30\% improvement over the weakest baselines under noisy conditions. The source code is available at https://github.com/LDer66/FEDRGL-ICML26.

Graphs are widely used to represent real-world relationships, such as social networks, recommendation systems, financial transactions, and healthcare records. In many applications, these graph data are distributed across different organizations or users and cannot be centralized due to privacy or ownership concerns. Federated graph learning allows multiple parties to collaboratively train models without sharing their raw graph data, but real-world graph labels are often noisy or incorrect, which can seriously affect model reliability. This paper proposes FedRGL, a robust federated graph learning method for clients holding partial graph data with noisy labels. FedRGL identifies unreliable labeled nodes by combining information from the global model and each client’s local graph structure, recovers useful supervision from some noisy nodes through high-confidence pseudo-labels, and reduces the influence of unreliable client updates during model aggregation. Experiments on multiple graph datasets show that FedRGL consistently improves performance under different types and levels of label noise. This work can help build more reliable distributed graph learning systems for privacy-sensitive and decentralized applications.