DemonAgent：针对基于 LLM 的智能体的动态加密多后门植入攻击
DemonAgent: Dynamically Encrypted Multi-Backdoor Implantation Attack on LLM-based Agent
AI总结：
针对基于LLM的智能体后门易被安全审计发现的问题，提出动态加密多后门植入攻击方法，将后门映射为良性内容并分解为子片段以规避审计，在多数据集上实现近100%攻击成功率与0%检测率。
摘要：
随着基于 LLM 的智能体日益普及，后门可通过用户查询或环境反馈植入智能体中，引发了关于安全漏洞的严重担忧。然而，后门攻击通常会被分析智能体推理过程的安全审计所发现。为此，我们提出了一种名为动态加密多后门植入攻击的新型后门植入策略。具体而言，我们引入了动态加密，将后门映射为良性内容，从而有效规避安全审计。为了增强隐蔽性，我们进一步将后门分解为多个子后门片段。基于这些改进，后门能够显著地绕过安全审计。此外，我们提出了 AgentBackdoorEval，一个旨在全面评估智能体后门攻击的数据集。跨多个数据集的实验结果表明，我们的方法实现了接近 100% 的攻击成功率，同时保持 0% 的检测率，证明了其在规避安全审计方面的有效性。我们的发现突显了现有安全机制在检测高级攻击方面的局限性，强调了针对后门威胁需要更强大防御的迫切需求。代码和数据可在 https://github.com/whfeLingyu/DemonAgent 获取。
英文摘要：
As LLM-based agents become increasingly prevalent, backdoors can be implanted into agents through user queries or environment feedback, raising critical concerns regarding safety vulnerabilities. However, backdoor attacks are typically detectable by safety audits that analyze the reasoning process of agents. To this end, we propose a novel backdoor implantation strategy called \textbf{Dynamically Encrypted Multi-Backdoor Implantation Attack}. Specifically, we introduce dynamic encryption, which maps the backdoor into benign content, effectively circumventing safety audits. To enhance stealthiness, we further decompose the backdoor into multiple sub-backdoor fragments. Based on these advancements, backdoors are allowed to bypass safety audits significantly. Additionally, we present AgentBackdoorEval, a dataset designed for the comprehensive evaluation of agent backdoor attacks. Experimental results across multiple datasets demonstrate that our method achieves an attack success rate nearing 100\% while maintaining a detection rate of 0\%, illustrating its effectiveness in evading safety audits. Our findings highlight the limitations of existing safety mechanisms in detecting advanced attacks, underscoring the urgent need for more robust defenses against backdoor threats. Code and data are available at https://github.com/whfeLingYu/DemonAgent.
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。
