Chenqing Zhu

Chenqing Zhu, Yanbo Dai, Yulong Tian et al.

Large Language Model (LLM)-based question-answering (QA) systems are increasingly deployed in sensitive domains such as healthcare, mental health counseling, and legal consultation. Federated learning (FL) enables collaborative training without sharing raw client data, for which locally trained models are aggregated at a central server (i.e., a cloud service provider) to obtain a global model. In this paper, we explore the potential vulnerability where a malicious aggregator, who may collude with a third-party vendor, stealthily implants advertisement-type backdoors into federated QA models, without ever accessing client data. The attacker's goals are twofold: (1) preserve clean QA fidelity (i.e., the poisoned model behaves like a clean model on non-triggered queries); and (2) generate highly natural, contextually relevant responses with target advertisements when a trigger appears. Achieving these two goals simultaneously is highly challenging, as naive backdoor injection without knowledge about private data may degrade model's clean performance or fail to inject the target. Motivated by this, we propose to leverage clients' uploaded gradients during training, and develop a two-stage framework for data-free and stealthy poisoning: (1) recover representative training samples from client gradients, and (2) construct poisoning datasets utilizing recovered samples and trigger phrases to inject backdoors into the global model. Experiments across representative QA datasets and LLM families under full fine-tuning and LoRA settings demonstrate that, our method achieves nearly 100% Attack Success Rate (ASR) while incurring negligible degradation on clean tasks. Crucially, reconstructing only 5-20% of gradients suffices to mount a reliable attack, exposing a practical blind spot in the pipeline of federated training of QA LLMs.

Songze Li, Chenqing Zhu

Conventional federated learning (FL) frameworks follow a server-driven model where the server determines session initiation and client participation, which faces challenges in accommodating clients' asynchronous needs for model updates. We introduce Client-Driven Federated Learning (CDFL), a novel FL framework that puts clients at the driving role. In CDFL, each client independently and asynchronously updates its model by uploading the locally trained model to the server and receiving a customized model tailored to its local task. The server maintains a repository of cluster models, iteratively refining them using received client models. Our framework accommodates complex dynamics in clients' data distributions, characterized by time-varying mixtures of cluster distributions, enabling rapid adaptation to new tasks with superior performance. In contrast to traditional clustered FL protocols that send multiple cluster models to a client to perform distribution estimation, we propose a paradigm that offloads the estimation task to the server and only sends a single model to a client, and novel strategies to improve estimation accuracy. We provide a theoretical analysis of CDFL's convergence. Extensive experiments across various datasets and system settings highlight CDFL's substantial advantages in model performance and computation efficiency over baselines.

Jin Xie, Chenqing Zhu, Songze Li

We focus on the problem of Personalized Federated Continual Learning (PFCL): a group of distributed clients, each with a sequence of local tasks on arbitrary data distributions, collaborate through a central server to train a personalized model at each client, with the model expected to achieve good performance on all local tasks. We propose a novel PFCL framework called Federated Memory Strengthening FedMeS to address the challenges of client drift and catastrophic forgetting. In FedMeS, each client stores samples from previous tasks using a small amount of local memory, and leverages this information to both 1) calibrate gradient updates in training process; and 2) perform KNN-based Gaussian inference to facilitate personalization. FedMeS is designed to be task-oblivious, such that the same inference process is applied to samples from all tasks to achieve good performance. FedMeS is analyzed theoretically and evaluated experimentally. It is shown to outperform all baselines in average accuracy and forgetting rate, over various combinations of datasets, task distributions, and client numbers.