ByzFL: Research Framework for Robust Federated Learning

TL;DR

ByzFL is an open-source Python library that streamlines the development and benchmarking of robust federated learning algorithms through a unified framework supporting various attacks, scenarios, and visualization tools.

Contribution

It introduces a comprehensive, extensible framework for robust federated learning research, integrating state-of-the-art algorithms, attack simulations, and scenario configurations.

Findings

Facilitates systematic experimentation with federated learning scenarios

Supports benchmarking of robust aggregators against adversarial attacks

Enhances reproducibility and rapid prototyping in FL research

Abstract

We present ByzFL, an open-source Python library for developing and benchmarking robust federated learning (FL) algorithms. ByzFL provides a unified and extensible framework that includes implementations of state-of-the-art robust aggregators, a suite of configurable attacks, and tools for simulating a variety of FL scenarios, including heterogeneous data distributions, multiple training algorithms, and adversarial threat models. The library enables systematic experimentation via a single JSON-based configuration file and includes built-in utilities for result visualization. Compatible with PyTorch tensors and NumPy arrays, ByzFL is designed to facilitate reproducible research and rapid prototyping of robust FL solutions. ByzFL is available at https://byzfl.epfl.ch/, with source code hosted on GitHub: https://github.com/LPD-EPFL/byzfl.