Robust Knowledge Distillation in Federated Learning: Counteracting Backdoor Attacks

TL;DR

This paper introduces Robust Knowledge Distillation (RKD), a new defense method for federated learning that effectively detects and mitigates backdoor attacks without relying on restrictive assumptions, ensuring model integrity and performance.

Contribution

RKD is a novel defense mechanism combining clustering, model selection, and knowledge distillation to improve backdoor attack resistance in federated learning.

Findings

RKD outperforms existing defenses across multiple scenarios.

RKD maintains high model accuracy while defending against backdoors.

RKD effectively filters malicious updates without strict data assumptions.

Abstract

Federated Learning (FL) enables collaborative model training across multiple devices while preserving data privacy. However, it remains susceptible to backdoor attacks, where malicious participants can compromise the global model. Existing defence methods are limited by strict assumptions on data heterogeneity (Non-Independent and Identically Distributed data) and the proportion of malicious clients, reducing their practicality and effectiveness. To overcome these limitations, we propose Robust Knowledge Distillation (RKD), a novel defence mechanism that enhances model integrity without relying on restrictive assumptions. RKD integrates clustering and model selection techniques to identify and filter out malicious updates, forming a reliable ensemble of models. It then employs knowledge distillation to transfer the collective insights from this ensemble to a global model. Extensive evaluations demonstrate that RKD effectively mitigates backdoor threats while maintaining high model performance, outperforming current state-of-the-art defence methods across various scenarios.