Zichuan Li (University of Illinois Urbana-Champaign), Jian Cui (University of Illinois Urbana-Champaign), Xiaojing Liao (University of Illinois Urbana-Champaign), Luyi Xing (University of Illinois Urbana-Champaign)

Large Language Model (LLM) agents are autonomous systems powered by LLMs, capable of reasoning and planning to solve problems by leveraging a set of tools. However, the integration of multiple tools in LLM agents introduces challenges in securely managing tools, ensuring their compatibility, handling dependency relationships, and protecting control flows within LLM agent's task workflows. In this paper, we present the first systematic security analysis of task control flows in multi-tool-enabled LLM agents. We identify a novel threat, Cross-Tool Harvesting and Polluting (XTHP), which includes multiple attack vectors to first hijack the normal control flows of agent tasks, and then collect and pollute confidential or private information within LLM agent systems. To understand the impact of this threat, we developed Chord, a dynamic scanning tool designed to automatically detect real-world agent tools susceptible to XTHP attacks. Our evaluation of 66 real-world tools from two major LLM agent development frameworks, LangChain and Llama-Index, revealed that 75% are vulnerable to XTHP attacks, highlighting the prevalence of this threat.

View More Papers

Are your Sites Truly Isolated? Automatically Detecting Logic Bugs...

Jan Drescher (TU Braunschweig), David Klein (TU Braunschweig), Martin Johns (TU Braunschweig)

Read More

Towards automated threat modeling for space systems via SPARTA...

Joonhyuk Park (School of Cybersecurity, Korea University), Jiwon Kwak (School of Cybersecurity, Korea University), Geunwoo Baek (School of Cybersecurity, Korea University), Dohee Kang (School of Cybersecurity, Korea University), Seungjoo Kim (School of Cybersecurity, Korea University)

Read More

Enhancing Website Fingerprinting Attacks against Traffic Drift

Xinhao Deng (INSC, Tsinghua University and Ant Group), Yixiang Zhang (INSC, Tsinghua University), Qi Li (INSC, Tsinghua University, State Key Laboratory of Internet Architecture, Tsinghua University and Zhongguancun Laboratory), Zhuotao Liu (INSC, Tsinghua University and Zhongguancun Laboratory), Yabo Wang (DCST, Tsinghua University), Ke Xu (DCST, Tsinghua University, State Key Laboratory of Internet Architecture, Tsinghua University…

Read More