Chunyi Zhou

FedGraph: Defending Federated Large Language Model Fine-Tuning Against Backdoor Attacks via Graph-Based Aggregation

Xi Chen, Chunyi Zhou, Rui Zeng +3

Federated fine-tuning of large language models (LLMs) allows multiple clients to collaboratively adapt models without sharing raw data, offering strong privacy benefits. However, t…

Safety in Self-Evolving Agents: A Survey

Jiahao Chen, Zhou Feng, Oubo Ma +28

Large language models (LLMs) exhibit strong general capabilities, yet their parameters typically remain fixed after deployment, limiting learning from new interactions. In open-end…

A Finger on the Scale: Covert Policy Steering through Agentic Skills

Jiarui Li, Jiahao Chen, Chunyi Zhou +5

Reusable agent skills extend large language model (LLM) agents with task procedures, tool-use guidance, and output constraints. Yet these skills also act as externalized behavioral…

The Boy Who Cried Wolf: Adversarial Misclassification of Safe Inputs as Unsafe in Multimodal Guardrails

Shuo Shi, Rui Yin, Naen Xu +7

Multimodal guard models have emerged as critical safety components for screening content in vision-language systems. While adversarial research has extensively studied jailbreaking…

Lilith: Backdoor Generalization under Training-Inference Trigger Shift

Zhou Feng, Jiahao Chen, Chunyi Zhou +6

Machine-learning services increasingly rely on public data, third-party providers, and outsourced training, creating opportunities for data-poisoning attacks that implant persisten…

Customization under Fire: Plugin Poisoning in Text-to-Image Ecosystem

Jiahao Chen, Xing He, Yong Yang +6

The prosperity of text-to-image (T2I) models has fostered a vibrant share-and-play ecosystem centered on Low-Rank Adaptation (LoRA) plugins, which allow users to customize and shar…

Profiling for Pennies: Unveiling the Privacy Iceberg of LLM Agents

Jiahao Chen, Qi Zhang, Ruixiao Lin +7

Large Language Models (LLMs) have revolutionized how information are collected, aggregated, and reasoned. However, this enables a novel and accessible vector of privacy intrusion:…

Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors

Rui Yin, Tianxu Han, Naen Xu +8

Safety-aligned large language models (LLMs) are increasingly deployed in real-world pipelines, yet this deployment also enlarges the supply-chain attack surface: adversaries can di…

Unveiling the Security Risks of Federated Learning in the Wild: From Research to Practice

Jiahao Chen, Zhiming Zhao, Yuwen Pu +4

Federated learning (FL) has attracted substantial attention in both academia and industry, yet its practical security posture remains poorly understood. In particular, a large body…

FraudShield: Knowledge Graph Empowered Defense for LLMs against Fraud Attacks

Naen Xu, Jinghuai Zhang, Ping He +6

Large language models (LLMs) have been widely integrated into critical automated workflows, including contract review and job application processes. However, LLMs are susceptible t…

DP-GENG : Differentially Private Dataset Distillation Guided by DP-Generated Data

Shuo Shi, Jinghuai Zhang, Shijie Jiang +5

Dataset distillation (DD) compresses large datasets into smaller ones while preserving the performance of models trained on them. Although DD is often assumed to enhance data priva…

Auditing M-LLMs for Privacy Risks: A Synthetic Benchmark and Evaluation Framework

Junhao Li, Jiahao Chen, Zhou Feng +1

Recent advances in multi-modal Large Language Models (M-LLMs) have demonstrated a powerful ability to synthesize implicit information from disparate sources, including images and t…

FreeTalk:A plug-and-play and black-box defense against speech synthesis attacks

Yuwen Pu, Zhou Feng, Chunyi Zhou +4

Recently, speech assistant and speech verification have been used in many fields, which brings much benefit and convenience for us. However, when we enjoy these speech applications…

IPIGuard: A Novel Tool Dependency Graph-Based Defense Against Indirect Prompt Injection in LLM Agents

Hengyu An, Jinghuai Zhang, Tianyu Du +4

Large language model (LLM) agents are widely deployed in real-world applications, where they leverage tools to retrieve and manipulate external data for complex tasks. However, whe…

LoRAShield: Data-Free Editing Alignment for Secure Personalized LoRA Sharing

Jiahao Chen, Junhao Li, Yiming Wang +6

The proliferation of Low-Rank Adaptation (LoRA) models has democratized personalized text-to-image generation, enabling users to share lightweight models (e.g., personal portraits)…

Poison in the Well: Feature Embedding Disruption in Backdoor Attacks

Zhou Feng, Jiahao Chen, Chunyi Zhou +3

Backdoor attacks embed malicious triggers into training data, enabling attackers to manipulate neural network behavior during inference while maintaining high accuracy on benign in…

"No Matter What You Do": Purifying GNN Models via Backdoor Unlearning

Jiale Zhang, Chengcheng Zhu, Bosen Rao +5

Recent studies have exposed that GNNs are vulnerable to several adversarial attacks, among which backdoor attack is one of the toughest. Similar to Deep Neural Networks (DNNs), bac…

CAMH: Advancing Model Hijacking Attack in Machine Learning

Xing He, Jiahao Chen, Yuwen Pu +5

In the burgeoning domain of machine learning, the reliance on third-party services for model training and the adoption of pre-trained models have surged. However, this reliance int…

Enhancing Adversarial Transferability with Adversarial Weight Tuning

Jiahao Chen, Zhou Feng, Rui Zeng +6

Deep neural networks (DNNs) are vulnerable to adversarial examples (AEs) that mislead the model while appearing benign to human observers. A critical concern is the transferability…

Rethinking the Vulnerabilities of Face Recognition Systems:From a Practical Perspective

Jiahao Chen, Zhiqiang Shen, Yuwen Pu +5

Face Recognition Systems (FRS) have increasingly integrated into critical applications, including surveillance and user authentication, highlighting their pivotal role in modern se…