arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2604.03081cs.CRcs.AIcs.CL
收藏

针对LLM编码代理技能生态系统的供应链污染攻击

Supply-Chain Poisoning Attacks Against LLM Coding Agent Skill Ecosystems

Griffith University(格里菲斯大学);The State Information Center(国家信息中心);Nanyang Technological University(南洋理工大学);University of New South Wales(新南威尔士大学);Wake Forest University(维克森林大学)

Yubin Qu, Yi Liu, Tongcheng Geng, Gelei Deng, Yuekang Li, Leo Yu Zhang, Ying Zhang, Lei Ma

AI总结:

本文研究了LLM编码代理技能生态系统中供应链攻击对行动空间的影响,提出DDIPE方法通过嵌入恶意逻辑到技能文档中实现隐式payload执行,实验显示其在强防御下可绕过11.6%-33.5%。

摘要:

基于LLM的编码代理通过第三方技能扩展能力,这些技能通过开放市场分发且无强制安全审查。与传统包不同,这些技能作为操作指令执行,具有系统级权限,因此单个恶意技能可破坏主机。现有研究未探讨供应链攻击能否直接劫持代理的行动空间,如文件写入、shell命令和网络请求。本文提出Document-Driven Implicit Payload Execution (DDIPE),通过在技能文档中的代码示例和配置模板中嵌入恶意逻辑,使代理在正常任务中重用这些示例时自动执行payload。使用LLM驱动的管道,我们从81个种子生成1,070个对抗性技能,覆盖15个MITRE ATTACK类别。在四个框架和五个模型上,DDIPE实现11.6%至33.5%的绕过率,而显式指令攻击在强防御下为0%。静态分析检测大多数情况,但2.5%的案例逃避检测和对齐。负责任的披露导致四个确认漏洞和两个修复。

英文摘要:

LLM-based coding agents extend their capabilities via third-party agent skills distributed through open marketplaces without mandatory security review. Unlike traditional packages, these skills are executed as operational directives with system-level privileges, so a single malicious skill can compromise the host. Prior work has not examined whether supply-chain attacks can directly hijack an agent's action space, such as file writes, shell commands, and network requests, despite existing safeguards. We introduce Document-Driven Implicit Payload Execution (DDIPE), which embeds malicious logic in code examples and configuration templates within skill documentation. Because agents reuse these examples during normal tasks, the payload executes without explicit prompts. Using an LLM-driven pipeline, we generate 1,070 adversarial skills from 81 seeds across 15 MITRE ATTACK categories. Across four frameworks and five models, DDIPE achieves 11.6% to 33.5% bypass rates, while explicit instruction attacks achieve 0% under strong defenses. Static analysis detects most cases, but 2.5% evade both detection and alignment. Responsible disclosure led to four confirmed vulnerabilities and two fixes.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑