Fedstellar: A Platform for Decentralized Federated Learning

TL;DR

Fedstellar addresses the need for flexible, privacy-preserving machine learning across heterogeneous IoT-like federations by enabling real-time, decentralized model training and aggregation. It extends the p2pfl framework into a modular platform that supports DFL, SDFL, and CFL architectures, with a Python-based core, a web-based frontend, and a controller to orchestrate deployments on physical and virtual devices. The authors validate Fedstellar through physical deployments on Raspberry Pi and Rock64 devices and virtualized experiments with MNIST and CIFAR-10, demonstrating robust performance (e.g., $F_{1}$ scores around $0.91$–$0.98$) and reduced training time compared to centralized baselines, while supporting diverse topologies and aggregation algorithms. The work highlights practical implications for scalable, resilient FL in IoT environments, offering a flexible platform for researchers to explore topology-aware, asynchronous, and secure federated learning at the edge.

Abstract

In 2016, Google proposed Federated Learning (FL) as a novel paradigm to train Machine Learning (ML) models across the participants of a federation while preserving data privacy. Since its birth, Centralized FL (CFL) has been the most used approach, where a central entity aggregates participants' models to create a global one. However, CFL presents limitations such as communication bottlenecks, single point of failure, and reliance on a central server. Decentralized Federated Learning (DFL) addresses these issues by enabling decentralized model aggregation and minimizing dependency on a central entity. Despite these advances, current platforms training DFL models struggle with key issues such as managing heterogeneous federation network topologies. To overcome these challenges, this paper presents Fedstellar, a platform extended from p2pfl library and designed to train FL models in a decentralized, semi-decentralized, and centralized fashion across diverse federations of physical or virtualized devices. The Fedstellar implementation encompasses a web application with an interactive graphical interface, a controller for deploying federations of nodes using physical or virtual devices, and a core deployed on each device which provides the logic needed to train, aggregate, and communicate in the network. The effectiveness of the platform has been demonstrated in two scenarios: a physical deployment involving single-board devices such as Raspberry Pis for detecting cyberattacks, and a virtualized deployment comparing various FL approaches in a controlled environment using MNIST and CIFAR-10 datasets. In both scenarios, Fedstellar demonstrated consistent performance and adaptability, achieving F1 scores of 91%, 98%, and 91.2% using DFL for detecting cyberattacks and classifying MNIST and CIFAR-10, respectively, reducing training time by 32% compared to centralized approaches.

Figures (10)

• Figure 1: Model training process in CFL and DFL
• Figure 2: Components of the Fedstellar platform: user, frontend, controller, and core
• Figure 3: Sequence diagram showing the interaction of Fedstellar components
• Figure 4: Overall architecture of the Fedstellar platform
• Figure 5: Types of deployment of Fedstellar: virtualized and physical scenarios