Another 215papers do not have category tags yet and are not included in filtered results.
Proposes LGWM, using an action-conditioned world model to verify GUI agent screen transitions.
提出 MFOTL 离线时序验证以监控工具 Agent 的危险动作链
这篇论文对工具调用大语言模型智能体的安全监控进行了形式化时序验证研究。
提出 ZoneClaw，用记忆分区防御计算机使用 Agent 的持久记忆攻击
提出 Tracekit，审计编码 Agent 的意图、推理与动作并检测账本篡改
Tracekit 是无第三方依赖的编码智能体审计层：把人类请求、模型自述和已执行动作并排写入可外部锚定的哈希链，执行前用正则门，事后用规则和独立模型互查。
构建 ACE 语料，评测内核级证据对 Agent 攻击检测的有效性
ACE 是一份把智能体会话的内核 syscall 痕迹与应用层清单、转录配对起来的检测语料和基准。。
用 FARSIGHT 方案级评测金融 LLM 交易 Agent 的稳健性与安全失败
提出 PMPA，经外部源向 Agent 持久记忆投毒并跨会话泄露隐私
Abstract论文提出了持久记忆投毒攻击 PMPA，报告了 harness 智能体在处理外部数据时易受跨会话记忆投毒与隐私泄露的风险。
提出类型化溯源与断言护栏，约束持久 Agent 释放自传断言
系统研究 LLM 智能体记忆投毒并构建基准 MPBench
Abstract系统研究了智能体记忆投毒风险，揭示了写入渠道漏洞，构建了 MPBench 基准并指出提示注入防御的局限。
提出潜伏记忆投毒，诱导 Agent 把外部文档写成持久伪造记忆
sleeper memory poisoning 研究带持久记忆的 LLM 助手：外部文档能否让助手存下一则伪造的用户记忆，并在原文档消失后的新会话里被取回、用来改变回复或动作。
提出 eTAMP，用网页内容跨会话投毒 Web Agent 轨迹记忆并诱导越权操作
该论文评估了基于大语言模型的 Web 智能体在利用历史交互轨迹进行记忆增强时的安全脆弱性。
Reviews threats to autonomous agents and proposes a full-lifecycle defense blueprint, FASA
This is a threat review and defense architecture report using OpenClaw as a case study, not a scored attack or defense evaluation.
Proposes a defense-in-depth aligned with the five stages to mitigate cross-stage threats to autonomous agents
This is a lifecycle threat analysis of OpenClaw, accompanied by a set of layered mitigation designs. The problem addressed is: when autonomous agents connect instant messaging, persistent memory, third-party plugins, and high-privilege execution, attacks are no longer limited to single-turn jailbreaks but can compound across multiple stages from initialization to execution. The authors argue that defenses hardening only individual interfaces cannot stop such cross-temporal attacks.
Proposes inference-time defense ICON, which detects attention over-focusing and redirects to correct indirect prompt injection in agents.
ICON detects attention over-focusing from IPI at inference time and redirects, avoiding refusal that would terminate the agent trajectory.
提出 Zombie Agent 攻击，经网页间接注入将载荷写入长期记忆并跨会话触发
Zombie Agent 把自进化 LLM 智能体的跨会话记忆当成攻击面：攻击者只发布网页，良性任务读入后，正常记忆更新留下载荷，并在以后无关会话里触发未授权工具行为。
提出 PASB，端到端评测 OpenClaw 的系统级攻击