Skip to content

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

Oct 13 · Tuesday

All papers

Found 24 papers

Another 201papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2610.09027

    P-VMI:对抗图像经 KV cache 持续影响多轮对话

    提出持久视觉记忆注入,使对抗图像经缓存在触发时操控输出

    Published
    Evaluated models
    Qwen3-8B, Qwen3-VL-8B-Instruct, Gemma-4-12B

    Abstract论文提出 P-VMI,证实对抗输入被掩蔽后仍可经 KV 缓存持久控制模型,强调了保留状态的安全风险。

    Paper analysisRead analysis
  2. MMSkillRisk:多模态技能图像中的隐藏指令可诱导 Agent 执行未授权操作

    提出 MMSkillRisk 基准,评测多模态技能图像中的隐藏指令诱导 Agent 越权操作

    Published
    Evaluated models
    GPT-5.6-sol, Kimi-K2.6, Kimi-K3 and others6

    Abstract论文提出 MMSkillRisk 基准与 NCVA 攻击,揭示多模态技能中教学图像潜藏指令可引发未授权操作并与良性任务共存。

    Paper analysisRead analysis
  3. AttacksarXiv 2609.37468

    论文提出针对 Agentic RAG 的检索器后门攻击与 DOU 隐藏方法

    提出检索器后门攻击操纵智能体搜索,并用 DOU 逃避检测

    Published
    Setting
    AI Agent
    Evaluated models
    E5, Contriever-MSMARCO

    Abstract论文展示了通过恶意检索器操纵智能体多轮搜索并利用诱饵重塑表征隐蔽后门的可行性与系统风险。

    Paper analysisRead analysis
  4. AttacksarXiv 2609.07048

    FreqDoor:面向视觉语言模型的频域后门攻击

    提出 FreqDoor,在频域混合振幅并对生成式 VLM 做训练时视觉后门

    Published
    Evaluated models
    BLIP-2 (opt-2.7b), InstructBLIP (flan-t5-xl), LLaVA (interleave-qwen)
    AbstractFreqDoor 用频域振幅混合给三款 VLM 植入仅视觉的训练时后门。

    FreqDoor 是针对生成式 VLM 的仅视觉、训练时后门:在频域混合触发源振幅、保留干净图像相位,使触发在空间上分散。

    Paper analysisRead analysis
  5. SinoGlyphBench:中文字形混淆下语言模型审核的诊断基准

    提出 SinoGlyphBench,用配对字形混淆诊断中文审核稳定性

    Published
    Evaluated models
    GPT-5.4 mini, GPT-5.5, DeepSeek V4 Flash and others12
    AbstractSinoGlyphBench 显示中文字形混淆会同时增加有害漏报和误报,全文扰动最重。

    SinoGlyphBench是一个配对诊断基准,用来看内容审核在可还原的中文字形混淆下是否还保持原决定。

    Paper analysisRead analysis
  6. AttacksarXiv 2609.02774

    CodePoisonRAG:针对检索增强代码生成的知识投毒攻击

    提出 CodePoisonRAG,对检索增强代码生成的知识库投毒以植入漏洞

    Published
    Evaluated models
    Qwen 3.5 9B, Code Llama 13B, DeepSeek-Coder-V2 16B

    AbstractCodePoisonRAG 展示了 RACG 外部知识库的新攻击面,在较低投毒率下可定向诱发指定漏洞并部分抵御防御。

    Paper analysisRead analysis
  7. AttacksarXiv 2606.01837

    论文提出 DSR 跨模态越狱框架:用良性输入诱导 MLLM 有害输出

    提出 DSR 跨模态越狱,用无害图文诱导 MLLM 生成有害图像

    Published
    Attacker
    Black-box
    Evaluated models
    gemini-2.5-flash-image, qwen-image-2.0, Tongyi Wanx 2.5
    AbstractDSR 用无害图文诱导商业 MLLM 输出有害图像,作者称组合能力会反过来扩大攻击面。

    DSR 是一种 black-box 跨模态越狱:输入侧不放显式有害内容,而让商业 MLLM 在推理中把无害图文重组成有害图像。

    Paper analysisRead analysis
  8. AttacksarXiv 2605.00583

    Jailbreaking vision-language models through the visual modality

    提出四类视觉越狱,把有害意图编码进图像以绕过安全拒答

    Published
    Evaluated models
    GPT-5.2, Claude-Haiku-4.5, Gemini-3-Flash and others8
    Abstract四类视觉越狱在六个 VLM 上检验文本对齐能否挡住视觉意图,模型与攻击之间差异大。

    四类视觉越狱用来检验前沿 VLM 的文本安全训练能否挡住由图像传达的有害意图。。

    Paper analysisRead analysis
  9. AttacksarXiv 2603.17239

    LAAF: A Systematic Red-Teaming Method and Automated Attack Framework for LPCI Vulnerabilities in Agentic LLM Systems

    Proposes LAAF, which automatically attacks LPCI vulnerabilities in agent systems in stages

    Published
    Setting
    AI Agent
    Evaluated models
    gemini-2.0-flash-001, claude-3-haiku-20240307, meta-llama/llama-3.1-70b-instruct and others5
    AbstractLAAF does LPCI red teaming with taxonomy plus stage seeding.

    The authors state that LAAF is an automated red-teaming framework built specifically for LPCI in agentic LLMs. The issue to address is that encoded, delayed, and conditionally triggered payloads can persist in memory or RAG and re-enter across sessions, whereas Garak and PyRIT do not use the previous round's breakthrough results to seed the next stage across the six stages.

    Paper analysisRead analysis
  10. AttacksarXiv 2603.07590

    Model as Lego builder: assembling malicious content from harmless blocks using a semantic blueprint

    Propose StructAttack, which uses structural diagrams to induce visual models to assemble malicious content from harmless semantic slots.

    Published
    Attacker
    Black-box
    Evaluated models
    GPT-4o (1120), Gemini-2.0-Flash (001), Gemini-2.5-Flash and others6
    AbstractStructAttack decomposes a harmful query into harmless slots and draws them into a structure graph, inducing the LVLM to fill in harmful content under a black-box single query.

    StructAttack is a black-box, single-query jailbreak against LVLMs: it decomposes a harmful instruction into separate seemingly harmless slots, draws them into a structural diagram, and then makes the model reassemble the slots into harmful content.

    Paper analysisRead analysis
  11. AttacksarXiv 2602.15927

    研究者提出面向多轮对话的视觉记忆注入攻击 VMI

    提出 VMI,用扰动图像在多轮对话中潜伏并仅在触发话题输出预定消息

    Published
    Evaluated models
    Qwen2.5-VL-7B-Instruct, Qwen3-VL-8B-Instruct, LLaVA-OneVision-1.5-8B-Instruct and others5
    AbstractVMI 让扰动图像在多轮 LVLM 中潜伏,作者称触发话题出现后仍可输出预定消息。

    VMI 是面向多轮 LVLM 的定向图像攻击:第三方发布带小扰动的图像,良性用户把它当作普通输入。

    Paper analysisRead analysis
  12. AttacksarXiv 2602.15654

    Zombie Agent: Persistent Control of Self-Evolving LLM Agents via Self-Reinforcing Injection

    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

    Published
    Setting
    AI Agent
    Evaluated models
    Gemini-2.5-Flash, GLM-4.7-Flash
    Abstract记忆写入把一次网页间接注入变成跨会话控制。

    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。

    Paper analysisRead analysis
  13. AttacksarXiv 2602.10148

    Red-Teaming Multimodal Reasoning: Cross-Modal Entanglement Attacks Jailbreak Vision-Language Models

    Propose COMET cross-modal entanglement attack to jailbreak vision-language models

    Published
    Attacker
    Black-box
    Evaluated models
    GPT-4.1, GPT-4.1-mini, Gemini-2.5-Flash and others9
    AbstractCOMET performs black-box cross-modal entanglement jailbreak.

    COMET is a black-box jailbreak for VLMs with multimodal reasoning, used to red-team cross-modal safety alignment. The gap the authors address is: distributing malicious cues into fixed combinations of image and text cannot detect vulnerabilities that only appear in multi-step reasoning. The attack does not access gradients; it only constructs image and text inputs.

    Paper analysisRead analysis
  14. AttacksarXiv 2601.15698

    Beyond Visual Safety: Jailbreaking Multimodal Large Language Models to Generate Harmful Images with Semantically Irrelevant Inputs

    提出 BVS 图文越狱,用中和拼接图和诱导提示词让模型生成有害图像

    Published
    Evaluated models
    GPT-5 (12 January 2026 release), Gemini 1.5 Flash (15 January 2026 release)
    AbstractBVS 以重建后再生成诱导有害图像。

    BVS 用图文对诱导 MLLM 生成有害图像。作者认为,相对文本越狱和单模态图像诱导,视觉生成的安全审查仍不充分。

    Paper analysisRead analysis