Watch Out for the Lifespan: Evaluating Backdoor Attacks Against Federated Model Adaptation

2025-11-18

AI Summary

This work presents the first systematic investigation into how Low-Rank Adaptation (LoRA) affects the lifecycle of backdoor attacks in federated learning (FL) model adaptation. We design an experimental framework integrating distributed training simulation with multi-scenario backdoor injection to quantitatively analyze the relationship between LoRA rank size and backdoor persistence. Our results demonstrate that low-rank LoRA significantly prolongs backdoor survival time, enhancing both stealthiness and malicious impact—thereby exposing critical limitations of existing FL security evaluation methods under dynamic adaptation settings. To address this gap, we propose the first evaluation paradigm specifically tailored to backdoor lifecycle analysis in parameter-efficient fine-tuning (PEFT) contexts. We publicly release our codebase and benchmark suite, establishing novel evaluation standards and foundational insights for securing FL model adaptation against backdoor threats.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large models adaptation through Federated Learning (FL) addresses a wide range of use cases and is enabled by Parameter-Efficient Fine-Tuning techniques such as Low-Rank Adaptation (LoRA). However, this distributed learning paradigm faces several security threats, particularly to its integrity, such as backdoor attacks that aim to inject malicious behavior during the local training steps of certain clients. We present the first analysis of the influence of LoRA on state-of-the-art backdoor attacks targeting model adaptation in FL. Specifically, we focus on backdoor lifespan, a critical characteristic in FL, that can vary depending on the attack scenario and the attacker's ability to effectively inject the backdoor. A key finding in our experiments is that for an optimally injected backdoor, the backdoor persistence after the attack is longer when the LoRA's rank is lower. Importantly, our work highlights evaluation issues of backdoor attacks against FL and contributes to the development of more robust and fair evaluations of backdoor attacks, enhancing the reliability of risk assessments for critical FL systems. Our code is publicly available.

Problem

• Evaluating backdoor attack impacts on federated learning model adaptation
• Analyzing LoRA parameter efficiency's effect on backdoor lifespan persistence
• Developing robust evaluation frameworks for backdoor attacks in federated systems

Innovation

• Evaluates backdoor attacks in federated learning
• Analyzes LoRA rank impact on backdoor lifespan
• Proposes robust evaluation framework for FL security