FoggyTrust: Robust Federated Learning with Hierarchical Trust Networks

A new hierarchical framework called FoggyTrust improves the resilience of federated learning systems against malicious actors by distributing trust computation to local fog nodes, according to a preprint posted to arXiv on June 26, 2026 [1][2]. The work extends FLTrust, an existing Byzantine-robust aggregation method that relies on a trusted server-side root dataset to score client updates [1][2]. FoggyTrust introduces a two-level architecture that localizes those trust scores to fog nodes, which the authors argue allows the system to handle globally heterogeneous data while preserving robustness within locally homogeneous client groups [1][2]. The framework pairs local trust-based aggregation with heterogeneity-aware global optimizers such as FedAdam and SCAFFOLD to address both distribution mismatch in trust estimation and client drift across groups [1][2]. On benchmark tests, FoggyTrust posted its strongest gains in challenging heterogeneous settings. On the CIFAR-10 dataset under Krum and Trim attacks, the model achieved an over 50% improvement compared with FLTrust [1][2]. The researchers also evaluated the framework on a real-world safari dataset for distributed wildlife monitoring, a use case they describe as socially impactful and safety-critical [1][2]. The preprint appeared on arXiv, the open-access repository that hosts electronic preprints across disciplines including computer science, mathematics, and physics [6]. As of November 2024, the repository was receiving about 24,000 new articles per month, and it surpassed two million total articles at the end of 2021 [6]. Papers on arXiv are moderated but not peer-reviewed before posting [6]. The FoggyTrust manuscript was submitted to the Machine Learning section of the site [1]. Federated learning frameworks have drawn increasing attention for their ability to train models across decentralized data sources without centralizing sensitive information [1][2]. Byzantine-robust variants specifically target scenarios where some participating clients may be compromised or intentionally disruptive [1][2]. The FoggyTrust authors contend that their hierarchical approach offers a practical path for deploying robust federated learning in environments where data distributions vary widely across geographic or organizational boundaries [1][2].