近期模型上下文协议(MCP)的进步使大语言模型(LLMs)能够轻松调用外部工具,从而产生新的强大且工具增强的代理。然而,这种能力也引入了未被探索的攻击面,即恶意操控工具响应。现有针对MCP的间接提示注入技术存在部署成本高、语义不连贯或白盒要求重的问题,且易被新提出的防御措施检测。本文提出Tree structured Injection for Payloads (TIP),一种新的黑盒攻击,生成自然负载以可靠控制MCP增强代理,即使在防御下也能有效。技术上,将负载生成视为树状结构搜索问题,并通过我们的粗到细优化框架指导搜索。为稳定学习并避免局部最优,引入路径感知反馈机制,仅将高质量历史轨迹反馈给攻击者模型。框架进一步通过显式条件化搜索于可观察的防御信号并动态重新分配探索预算,以增强对防御转换的抵抗力。在四个主流LLMs上的大量实验显示,TIP在无防御环境下攻击成功率超过95%,所需查询次数比现有自适应攻击少一个数量级。针对四个代表性防御方法,TIP保持超过50%的有效性,并显著优于现有最先进攻击。通过在现实世界MCP系统中实施攻击,我们的结果揭示了MCP部署中的隐形但实用的威胁向量。我们还讨论了潜在的缓解方法以解决这一关键安全缺口。
英文摘要:
Recent advances in the Model Context Protocol (MCP) have enabled large language models (LLMs) to invoke external tools with unprecedented ease. This creates a new class of powerful and tool augmented agents. Unfortunately, this capability also introduces an under explored attack surface, specifically the malicious manipulation of tool responses. Existing techniques for indirect prompt injection that target MCP suffer from high deployment costs, weak semantic coherence, or heavy white box requirements. Furthermore, they are often easily detected by recently proposed defenses. In this paper, we propose Tree structured Injection for Payloads (TIP), a novel black-box attack which generates natural payloads to reliably seize control of MCP enabled agents even under defense. Technically, We cast payload generation as a tree structured search problem and guide the search with an attacker LLM operating under our proposed coarse-to-fine optimization framework. To stabilize learning and avoid local optima, we introduce a path-aware feedback mechanism that surfaces only high quality historical trajectories to the attacker model. The framework is further hardened against defensive transformations by explicitly conditioning the search on observable defense signals and dynamically reallocating the exploration budget. Extensive experiments on four mainstream LLMs show that TIP attains over 95% attack success in undefended settings while requiring an order of magnitude fewer queries than prior adaptive attacks. Against four representative defense approaches, TIP preserves more than 50% effectiveness and significantly outperforms the state-of-the-art attacks. By implementing the attack on real world MCP systems, our results expose an invisible but practical threat vector in MCP deployments. We also discuss potential mitigation approaches to address this critical security gap.
Beyond Max Tokens: Stealthy Resource Amplification via Tool Calling Chains in LLM Agents
Nanyang Technological University, Singapore(南洋理工大学,新加坡);University of Illinois Urbana-Champaign, United States(伊利诺伊大学厄巴纳-香槟分校,美国);The Hong Kong University of Science and Technology, Hong Kong(香港科学与技术大学,香港);Shanghai Jiao Tong University, China(上海交通大学,中国)
Invisible to Humans, Triggered by Agents: Stealthy Jailbreak Attacks on Mobile Vision-Language Agents
School of Computer Science, Huazhong University of Science and Technology(华中科技大学计算机科学与技术学院);Dept. of Comp. Sci. and Tech., Institute for AI, Tsinghua-Bosch Joint ML Center, THBI Lab, BNRist Center, Tsinghua University(清华大学计算机科学与技术系、人工智能研究院、清华-博世机器学习联合中心、清华脑与智能实验室、北京国家信息科学与技术研究中心)
VisualTrap: A Stealthy Backdoor Attack on GUI Agents via Visual Grounding Manipulation
University of Science and Technology of China(中国科学技术大学);National University of Singapore(新加坡国立大学);East China University of Science and Technology(东华大学)
University of Pennsylvania(宾夕法尼亚大学);Brown University(布朗大学);Nanyang Technological University(南洋理工大学);University of Notre Dame(圣母大学);George Mason University(乔治梅森大学)
The Dark Side of LLMs: Agent-based Attack Vectors for System-level Compromise
DIMES , University of Calabria , P. Bucci , 87036 , Rende (CS) , Italy(DIMES,卡利博里大学,P. Bucci,87036,Rende(CS),意大利);IMT School for Advanced Studies , Piazza San Francesco , 55100 , Lucca , Italy(IMT高级研究学院,圣弗朗西斯科广场,55100,卢卡,意大利)
MIDS: Detecting Stealthy Masquerade and Tampering Attacks on CAN Bus via Bidirectional Mamba
Institute of Information Engineering, Chinese Academy of Sciences(信息工程研究所,中国科学院);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络安全学院);Zhongguancun Laboratory(中关村实验室)
GhostEncoder: Stealthy Backdoor Attacks with Dynamic Triggers to Pre-trained Encoders in Self-supervised Learning
Nanjing University of Aeronautics and Astronautics(南京航空航天大学);Nanjing University of Aeronautics and Astronautics Shenzhen Research Institute(南京航空航天大学深圳研究院);Zhejiang Lab(之江实验室);Wuhan University(武汉大学);Xi’an Jiaotong University(西安交通大学)
Tsinghua Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院);The State Key Laboratory of Blockchain and Data Security, Zhejiang University(浙江大学区块链与数据安全国家重点实验室);College of Computing and Data Science, Nanyang Technological University(南洋理工大学计算与数据科学学院);School of Computer Science, Wuhan University(武汉大学计算机学院);Tencent Data Platform(腾讯数据平台);Research Center of Artificial Intelligence, Peng Cheng Laboratory(鹏城实验室人工智能研究中心);School of Cyber Science and Engineering, Wuhan University(武汉大学网络安全学院)
Unicode TAG-Block Concealment of Tool-Metadata Payloads in the Model Context Protocol: An Approval-View Fidelity Gap Across Three Independent Server Implementations
Department of Computer Science(计算机科学系);AI and Media Analysis Lab(人工智能与媒体分析实验室)
Invisible for both Camera and LiDAR: Security of Multi-Sensor Fusion based Perception in Autonomous Driving Under Physical-World Attacks
University of Michigan(密歇根大学);NVIDIA Research(英伟达研究院);Arizona State University(亚利桑那州立大学);Inceptio(嬴彻科技);Baidu Research(百度研究院);National Engineering Laboratory of Deep Learning Technology and Application(深度学习技术及应用国家工程实验室);University of California, Irvine(加州大学欧文分校);University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)