5 papers
Merge Hijacking: Backdoor Attacks to Model Merging of Large Language Models
Zenghui Yuan, Yangming Xu, Jiawen Shi +2
Model merging for Large Language Models (LLMs) directly fuses the parameters of different models finetuned on various tasks, creating a unified model for multi-domain tasks. Howeve…
SafeAgent: Safeguarding LLM Agents via an Automated Risk Simulator
Xueyang Zhou, Weidong Wang, Lin Lu +7
Large Language Model (LLM)-based agents are increasingly deployed in real-world applications such as "digital assistants, autonomous customer service, and decision-support systems"…
Prompt Injection Attack to Tool Selection in LLM Agents
Jiawen Shi, Zenghui Yuan, Guiyao Tie +3
Tool selection is a key component of LLM agents. A popular approach follows a two-step process - \emph{retrieval} and \emph{selection} - to pick the most appropriate tool from a to…
BadToken: Token-level Backdoor Attacks to Multi-modal Large Language Models
Zenghui Yuan, Jiawen Shi, Pan Zhou +2
Multi-modal large language models (MLLMs) extend large language models (LLMs) to process multi-modal information, enabling them to generate responses to image-text inputs. MLLMs ha…
Poisoned-MRAG: Knowledge Poisoning Attacks to Multimodal Retrieval Augmented Generation
Yinuo Liu, Zenghui Yuan, Guiyao Tie +4
Multimodal retrieval-augmented generation (RAG) enhances the visual reasoning capability of vision-language models (VLMs) by dynamically accessing information from external knowled…