4 papers
The Guard That Cried Wolf: How Scary Words Make Agent Guardrails Refuse Legitimate Actions
Yingjie Zhang, Yuanbo Xie, Kai Chen
Agent guardrails are checks that approve or refuse each action before an LLM executes it. Sometimes they refuse requests that are genuinely safe. This over-safety blocks deployment…
What If Prompt Injection Never Left? Rethinking Agent Security through Cross-Session Stored Prompt Injection
Yuanbo Xie, Wenlei Zhu, Tianyun Liu +5
Modern agentic systems fundamentally reshape the security boundary of LLMs by introducing persistent system state including memories, filesystems, tools, and other long-lived conte…
Detecting RAG Extraction Attack via Dual-Path Runtime Integrity Game
Yuanbo Xie, Yingjie Zhang, Yulin Li +5
Retrieval-Augmented Generation (RAG) systems augment large language models with external knowledge, yet introduce a critical security vulnerability: RAG Knowledge Base Leakage, whe…
Beyond Surface Alignment: Rebuilding LLMs Safety Mechanism via Probabilistically Ablating Refusal Direction
Yuanbo Xie, Yingjie Zhang, Tianyun Liu +2
Jailbreak attacks pose persistent threats to large language models (LLMs). Current safety alignment methods have attempted to address these issues, but they experience two signific…