Some text is shown in its original language while the English translation is being prepared.
BackSource arXiv
Hanling Tian, Gengyu Zhang, Zeyang Sha, Jingying Wang, Yuhang Liu, Zhehao Huang, Kun Yang, Xiaolin Huang
Indexed by this site 2026-10-12 14:27 2 hr ago
Originally published Date not given
Selected 2026-10-12 15:10
Category Attacks arXiv · arXiv:2608.23471 · Hanling Tian, Gengyu Zhang, Zeyang Sha, Jingying Wang, Yuhang Liu, Zhehao Huang, Kun Yang, Xiaolin Huang · Indexed by this site 2026-10-12 14:27 · Originally publishedDate not given SelectedAttention55 InjecMEM:针对 LLM Agent 记忆系统的记忆注入攻击 InjecMEM: Memory Injection Attack on LLM Agent Memory Systems
AI reading guide
研究者提出 InjecMEM,一种只需一次交互、无需读取或编辑记忆存储即可向 LLM Agent 记忆系统注入记录的攻击,使后续针对目标主题的查询被导向预设输出。在 MemoryOS 上,其检索成功率最高 35.4%、条件攻击成功率 76.6%,而 DPI、BadChain 与 GCG 基线均降至 0;攻击在良性记忆漂移下仍有效,非目标查询基本不受影响,在 MemGPT 上同样有效。同族联合优化可让命令迁移到 3B、14B 及微调变体,跨族迁移有限,但拼接多个单模型命令可同时影响 Qwen、Mistral 与 Llama。检索期防御测试显示 LLM-as-a-Judge、ProtectAI、PromptGuard 只能降低而不能消除投毒检索,困惑度过滤则以大量良性拦截为代价。论文还展示经被入侵工具写入记忆的间接注入,即使工具被修复,记录仍留在记忆中。
Deep reading Generating
Why it was selected
论文给出单次交互污染 Agent 记忆的攻击方法与成功率,部署记忆系统的团队可据此检查检索环节的暴露面。
Read original arxiv.org Why it was selected SelectedAttention55
Attention 0–100: content value contributes up to 70 and circulation up to 30; circulation decays over time. How it was rated
论文给出单次交互污染 Agent 记忆的攻击方法与成功率,部署记忆系统的团队可据此检查检索环节的暴露面。