Detecting and Mitigating Backdoor Attacks in OTA-FL Systems: A Two-Stage Robust Aggregation Scheme

TL;DR

This paper introduces a two-stage robust aggregation scheme for OTA-FL systems that detects and mitigates backdoor attacks by assigning trust scores and categorizing clients, effectively defending against stealthy malicious updates.

Contribution

The paper proposes a novel two-stage framework combining trust scoring and layered inspection tailored for OTA-FL to enhance backdoor attack resilience.

Findings

Effectively suppresses various backdoor attacks.

Maintains competitive main-task accuracy.

Improves detection of stealthy malicious updates.

Abstract

Over-the-air federated learning (OTA-FL) improves communication efficiency by exploiting the superposition property of wireless channels, but this same property also creates a critical security vulnerability: the parameter server (PS) cannot access individual local updates, making it difficult to identify and exclude poisoned gradients. The challenge is further exacerbated under non-independent and identically distributed (Non-IID) training data, where benign gradient drift can closely resemble malicious updates. In this paper, we propose a two-stage robust aggregation framework for defending against backdoor attacks in OTA-FL. Under our scheme, each client is first assigned a modality-aware multi-indicator trust score, where the specific indicators are selected according to the data modality (e.g., waveform, text, image) and model architecture to capture the most discriminative footprint of backdoor updates. Based on this score, the PS then performs trust-based multiple access (TBMA) to separate clients into trusted, suspicious, and malicious categories. Suspicious clients are further examined through PS-side layer-wise inspection and a longitudinal reputation mechanism. Experimental results on several datasets demonstrate that the proposed methodology effectively suppresses stealthy backdoor attacks, including bounded-scaling attacks, Euclidean-constrained attacks, Cosine-constrained attacks, and Neurotoxin, while maintaining competitive main-task accuracy.