Blades: A Simulator for Attacks and Defenses in Federated Learning

Federated learning enables distributed training across a set of clients, without requiring any of the participants to reveal their private training data to a centralized entity or each other. Due to the nature of decentralized execution, federated learning is vulnerable to attacks from adversarial (Byzantine) clients by modifying the local updates to their desires. Therefore, it is important to develop robust federated learning algorithms that can defend Byzantine clients without losing model convergence and performance. In the study of robustness problems, a simulator can simplify and accelerate the implementation and evaluation of attack and defense strategies. However, there is a lack of open-source simulators to meet such needs. Herein, we present Blades, a scalable, extensible, and easily configurable simulator to assist researchers and developers in efficiently implementing and validating novel strategies against baseline algorithms in robust federated learning. Blades is built upon a versatile distributed framework Ray, making it effortless to parallelize single machine code from a single CPU to multi-core, multi-GPU, or multi-node with minimal configurations. Blades contains built-in implementations of representative attack and defense strategies and provides user-friendly interfaces to easily incorporate new ideas. We maintain the source code and documents at https://github.com/bladesteam/blades.