Federated Behavioural Planes: Explaining the Evolution of Client Behaviour in Federated Learning

TL;DR

This paper introduces Federated Behavioural Planes (FBPs), a novel method for analyzing, visualizing, and explaining client behaviour in federated learning, improving understanding, security, and robustness of FL systems.

Contribution

The paper presents FBPs as a new approach to analyze client dynamics in FL, and introduces Federated Behavioural Shields to detect malicious clients, enhancing security.

Findings

FBPs effectively visualize client behaviour trajectories.

FBPs enable clustering of clients based on behaviour patterns.

Federated Behavioural Shields improve detection of malicious clients.

Abstract

Federated Learning (FL), a privacy-aware approach in distributed deep learning environments, enables many clients to collaboratively train a model without sharing sensitive data, thereby reducing privacy risks. However, enabling human trust and control over FL systems requires understanding the evolving behaviour of clients, whether beneficial or detrimental for the training, which still represents a key challenge in the current literature. To address this challenge, we introduce Federated Behavioural Planes (FBPs), a novel method to analyse, visualise, and explain the dynamics of FL systems, showing how clients behave under two different lenses: predictive performance (error behavioural space) and decision-making processes (counterfactual behavioural space). Our experiments demonstrate that FBPs provide informative trajectories describing the evolving states of clients and their contributions to the global model, thereby enabling the identification of clusters of clients with similar behaviours. Leveraging the patterns identified by FBPs, we propose a robust aggregation technique named Federated Behavioural Shields to detect malicious or noisy client models, thereby enhancing security and surpassing the efficacy of existing state-of-the-art FL defense mechanisms. Our code is publicly available on GitHub.