Skip to content

Translation in progress Some entries still include text in the original language.

Oct 12 · Monday

All papers

Found 21 papers

Another 240papers do not have category tags yet and are not included in filtered results.

  1. DefensesarXiv 2610.01349

    PACE:面向工具调用 LLM Agent 的来源感知能力强制执行

    提出 PACE,在工具调用前做路径隔离与效果授权

    Published
    Setting
    AI Agent
    Evaluated models
    gpt-5.6-luna, DeepSeek-V4-Flash, Qwen-3.8-27B
    AbstractPACE 在工具边界做路径割与效果核验。

    PACE 是面向工具调用型 LLM 智能体的运行时调解:当准入证据无法区分会改变行为的规格时,不在出口站点上撤销调解,而在最后一次工具调用前同时检查表示出的影响路径和请求权限。

    Paper analysisRead analysis
  2. DefensesarXiv 2610.02569

    Pincer:用数字分身学习用户偏好并执行动态最小权限

    为 Agent 提出资源层授权系统 Pincer,用数字分身执行动态最小权限

    Published
    Setting
    AI Agent
    Evaluated models
    Claude Haiku 4.5, Muse Spark 1.2 Contributor
    AbstractPincer 在资源层以三票合取做授权。

    Pincer 是加在现成 coding agent 上的资源层授权防御:用只受用户侧信息影响的 digital twin 做动态最小权限,并与 tool-call 层防御并存。要处理的是用户维护策略带来的权限蠕变和授权疲劳,以及不学习用户策略的 tool-call 分类器。作者引述工作称 auto mode 会放行多数不安全操作,并认为推理代码或复杂 shell 的效果在这一层不实际。

    Paper analysisRead analysis
  3. DefensesarXiv 2610.00371

    FlowReview:以授权配对评测控制多智能体组合信息流

    提出 FlowReview,用授权配对与确定性提交门控控制多智能体组合信息流

    Published
    Evaluated models
    Sonnet 4.5, Haiku 4.5, Qwen3-32B and others5
    Abstract配对标准要求拒绝禁止使用并完成授权使用,FlowReview 把三项审查接到执行。

    这篇工作把多智能体安全写成一件必须同时成立的事:拦住组合信息所启用的禁止使用,并完成同一场景里任务所需的授权使用。

    Paper analysisRead analysis
  4. AttacksarXiv 2609.27542

    Researchers use control token injection to delete gpt-oss-20b's reasoning chain and bypass chain-of-thought monitoring

    提出控制标记注入,清空工具 Agent 的思维链以绕过推理监控

    Published
    Setting
    AI Agent
    Evaluated models
    openai/gpt-oss-20b, google/gemma-4-26B-A4B-it, deepseek-ai/DeepSeek-R1-Distill-Qwen-7B and others4

    Abstract论文阐明控制标记注入可抑制 CoT 使监控失效并促成拒绝绕过,且工具执行受解析器宽容度影响。

    Paper analysisRead analysis
  5. DefensesarXiv 2609.22573

    研究提出面向企业 MCP 的零信任授权与工具发现方案

    提出企业 MCP 零信任授权与发现扩展,在架构上拒绝提示注入下的越权调用

    Published
    Setting
    AI Agent
    Evaluated models
    gpt-5, claude-sonnet-4-6, claude-opus-4-7 and others6
    Abstract同一份按工具声明同时管 MCP 的发现、列表和调用,使注入不能扩大已有凭证的权限。

    作者把零信任放在 MCP 工具调用边界:智能体即使被提示注入,也不能调用其凭证范围之外的工具。问题来自三点部署性质。调用意图会随不可信上下文改变。

    Paper analysisRead analysis
  6. DefensesarXiv 2609.17648

    多智能体 LLM 流水线中的信任传播与结构隔离

    用任务绑定令牌和策略预言机拦住多智能体流水线的未授权执行

    Published
    Evaluated models
    gemma4:31b-cloud
    Abstract结构性授权拦住了记忆投毒的未授权执行,持密钥时承重的是策略预言机。

    作者在一条会改写沙箱邮件库的四智能体流水线上,检查低权限失陷会不会变成未授权的真实动作。

    Paper analysisRead analysis
  7. DefensesarXiv 2609.14631

    IntentCap 提出按任务意图限定 LLM Agent 权限的机制

    提出 IntentCap,按来源字段所有权合成并裁决 Agent 能力租约

    Published
    Setting
    AI Agent
    Abstract四源字段租约约束副作用。

    IntentCap 是把智能体能力绑到当前任务意图、并按上下文来源划分字段所有权的权限机制。

    Paper analysisRead analysis
  8. OtherarXiv 2609.12413

    SoK 综述:智能体 AI 时代的越狱攻击、防御与实践考量

    系统梳理智能体执行链路中的越狱攻击与防御并比较推理时防御

    Published
    Setting
    AI Agent
    Evaluated models
    Qwen3.5-4B, Llama-3.1-8B-Instruct
    AbstractSoK 将越狱重放到智能体流水线,并用两种模型检验回复层防御的边际与代价。

    Mia 等人把越狱安全从单轮回复扩展到智能体的规划、记忆、工具和智能体间通信,并问既有结论在更强原生对齐下是否还成立。

    Paper analysisRead analysis
  9. DefensesarXiv 2609.29647

    AgentKernel:面向 Agent 的可信原生操作系统内核

    提出 AgentKernel,用四支柱强制中介 Agent 的输入、工具与记忆

    Published
    Setting
    AI Agent
    AbstractAgentKernel 是四支柱智能体 OS 设计,安全论证未附带实测数字。

    AgentKernel 是把安全做成组织原则的智能体 OS 架构,不是带跑分的攻防评测。

    Paper analysisRead analysis
  10. ActBench: A Self-Evolving Benchmark for Behavioral Safety of Collaborative Agents

    构建自演化基准 ActBench,评测协作智能体的行为安全

    Published
    Setting
    AI Agent
    Evaluated models
    Claude-Opus-4.8, Claude-Sonnet-4.6, GPT-5.5 and others15

    AbstractActBench 通过自演化搜索与双证据重构评测轨迹行为安全,揭示基础模型决策差异主导了协同智能体的操作风险。

    Paper analysisRead analysis
  11. AttacksarXiv 2605.08460

    论文建模多智能体网络中的子智能体继承攻击面

    建模并利用多智能体网络中子智能体生成时的记忆继承漏洞

    Published
    Evaluated models
    MiniMax M2.5, Llama-4-Maverick-17B-128E-Instruct-FP8, Qwen3.5-Plus and others5
    Abstract形式化子智能体继承,在 OpenClaw 上演示四类编排层漏洞并提出防御。

    作者对多智能体网络里的子智能体继承做形式化分析,并用开源框架上的 PoC 演示编排层如何让单点妥协继续扩散。

    Paper analysisRead analysis
  12. AttacksarXiv 2604.05549

    Stop obsessing over prompts: reasoning hijacking and constraint tightening for red-teaming LLM agents

    提出 JailAgent,不改用户提示,经推理劫持与记忆投毒操纵 Agent 动作

    Published
    Setting
    AI Agent
    Evaluated models
    GPT-3.5-turbo, GPT-4o, GPT-5 and others7
    AbstractJailAgent 不改提示词,用三阶段操纵推理与检索。

    JailAgent 是不修改用户原始提示词的 LLM 智能体红队框架。

    Paper analysisRead analysis
  13. DefensesarXiv 2603.12644

    Revealing security threats and defense architectures for autonomous agents: an OpenClaw case study

    Reviews threats to autonomous agents and proposes a full-lifecycle defense blueprint, FASA

    Published
    Setting
    AI Agent
    AbstractUsing OpenClaw as an example, it divides risks into three layers and uses FASA to describe defense from input to system execution.

    This is a threat review and defense architecture report using OpenClaw as a case study, not a scored attack or defense evaluation.

    Paper analysisRead analysis
  14. DefensesarXiv 2603.11619

    Taming OpenClaw: Security Analysis and Mitigation of Threats in Autonomous LLM Agents

    Proposes a defense-in-depth aligned with the five stages to mitigate cross-stage threats to autonomous agents

    Published
    Setting
    AI Agent
    AbstractThis is a lifecycle threat analysis and layered mitigation design for OpenClaw, with the cases serving as qualitative evidence.

    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.

    Paper analysisRead analysis