Skip to content

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

Oct 12 · Monday

All papers

Found 224 papers
Filter4

Attack type

Jailbreaks60Prompt injection7Poisoning and backdoors29Model tampering19Extraction and theft13Detection evasion19Malicious use10

Attack technique

Gradient and representation optimization23Automated search20Multi-turn escalation5Encoding and obfuscation20Role-play and persuasion3Cross-modal payloads44Defense-targeted10Dormant triggers17

Model risks

Reward hacking3Misalignment and value divergence3Inappropriate refusal9Hallucination and fabrication11

Defense type

Model hardening37Inference-time intervention13Detection and filtering14Monitoring and auditing5Watermarking and provenance5

Impact

Harmful content90Data and privacy leaks18Output manipulation70Availability and reliability5Oversight failure3

System type

Models and APIs220LLM applications15AI Agent42coding agent4web agent7Browser and computer use6Embodied systems and robotics15

Attacker access

Black-box47Gray-box5White-box21No-box (offline transfer)4

Attack surface / entry point

User input74Images, audio and video92Training data38Weights and gradients34Distribution channels12

Layer

Prompt layer85Model layer69Application layer4Training and data layer65

System component

MCP and skills24Memory11Guardrails and judges91Monitors48Reasoning chains62Permissions and sandboxing12Fine-tuning and open weights59Vision98Audio32Video11Image generation46Embeddings4Multimodal177
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 254papers have no tags yet and are not included in filtered results.
System type：Models and APIsSystem type：coding agentSystem component：Fine-tuning and open weightsSystem component：MultimodalClear all

Another 254papers do not have category tags yet and are not included in filtered results.

  1. DefensesarXiv 2610.13733

    KVeil: First-Place Solution for the ECCV 2026 LifeGenIP Unlearnable Video Challenge

    Propose KVeil, which adds perturbations to face videos to prevent unauthorized customization from reproducing identity

    arXiv
    2610.13733
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    White-box, Black-box
    Evaluated models
    Wan2.2-TI2V-5B, HunyuanVideo-1.5, Kling 3.0 and others4

    AbstractKVeil adapts LETHE's reference-reading disruption to the challenge's two stages, ranking first on the leaderboard in both Final columns.

    Paper analysisRead analysis
  2. DefensesarXiv 2610.13333

    LETHE: Defending Against Unauthorized Image-to-Video Generation via Value Diversity Collapse

    Proposes LETHE, which perturbs reference images to block unauthorized image-to-video

    arXiv
    2610.13333
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Wan2.1-I2V-14B, CogVideoX-5B-I2V, LightX2V and others6
    AbstractLETHE suppresses the diversity of reference values and attracts attention, making protected images difficult to animate faithfully.

    LETHE is a protective perturbation added to personal images before publication, used to counter unauthorized image-to-video generation.

    Paper analysisRead analysis
  3. AttacksarXiv 2610.12644

    Cancerous backdoor attack: progressive activation via invisible tokens and a stage controller.

    Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.

    arXiv
    2610.12644
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3
    AbstractHidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.

    Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.

    Paper analysisRead analysis
  4. Evaluation and benchmarksarXiv 2610.11112

    EpiReal-Bench：商用图像生成模型虚假声明红队基准

    构建 EpiReal-Bench，评测图像生成器将虚假主张渲染为可信证据

    arXiv
    2610.11112
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    GPT-Image-2, Nano Banana 2, Grok Imagen 2.0 and others4
    Abstract四款商用图像模型能把假主张画成可信证据。

    EpiReal-Bench 用来量商用图像生成器会不会把虚假现实主张画成可信视觉证据。

    Paper analysisRead analysis
  5. InterpretabilityarXiv 2610.10865

    用路由稀疏自编码器解耦语音编码器中的语言学与副语言学信息

    提出路由 TopK SAE，把语音编码器稀疏码拆成语言与副语言通路

    arXiv
    2610.10865
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    SPEAR XLarge, WavLM Large
    Abstract路由 SAE 在冻结 SPEAR 与 WavLM 上分开语言和副语言因子，探针与交换支持通路分工。

    TopK SAE 的稀疏码被分成语言通路与副语言通路，用来分开冻结语音编码器里同时存在的音素信息和说话人、情感、韵律。只重构不能得到作者选定的这一分解。语言通路由音素识别监督，副语言通路由说话人识别、情感分类和韵律预测监督，作用在对方通路上的梯度反转对抗用来压低交叉因子。

    Paper analysisRead analysis
  7. DefensesarXiv 2610.10345

    SLDR：通过选择性层恢复与动态路由防御恶意微调

    提出 SLDR，用选择性层恢复与动态路由防御恶意微调

    arXiv
    2610.10345
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Llama3.1-8B-Instruct, Llama3-8B-Instruct, Qwen2.5-7B-Instruct and others4
    Abstract作者称 SLDR 能压低有害输出并保持下游效用。

    SLDR 是针对恶意微调的微调后防御：在已对齐模型上做有符号层探测，只在敏感分数最大与最小的两层训练 LoRA，再用 lmax 的最后 token 表示决定是否启用适配器。

    Paper analysisRead analysis
  8. DefensesarXiv 2610.09941

    OrthoPurify 用一步正交投影净化被后门污染的视觉语言模型

    提出 OrthoPurify，用正交投影清除视觉语言模型适配器中的劫持方向

    arXiv
    2610.09941
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    LLaVA-1.5-7B, Qwen3-VL-8B, LLaVA-1.5-13B and others5
    Abstract一次投影去掉劫持方向。

    OrthoPurify 是部署前的 LVLM 后门净化方法，用一次正交投影去掉被劫持的权重更新方向。

    Paper analysisRead analysis
  9. AttacksarXiv 2610.09819

    FAB 攻击：给 HuBERT/WavLM 植入后门，背景警笛声即可让下游语音任务失灵

    提出 FAB 攻击，向声学基础模型植入可由环境声触发的后门

    arXiv
    2610.09819
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    HuBERT-base, WavLM-base

    Abstract论文提出了针对声学基模的 FAB 后门攻击，在弱威胁模型下实现了任务无关、物理可触发的后门植入。

    Paper analysisRead analysis
  10. DefensesarXiv 2610.09703

    研究揭示 Token-Pruning 对 VLM 安全性的影响并提出 SAP 缓解机制

    提出 Safety-Aware Pruning，缓解视觉 token 剪枝放大的多模态越狱

    arXiv
    2610.09703
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    LLaVA-1.5-7B
    Abstract剪枝去掉背景 token 后注意力塌向恶意前景。

    这篇工作评估视觉 token 剪枝如何改变 VLM 在多模态越狱下的行为，并给出推理时防御 SAP。

    Paper analysisRead analysis
  11. 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

    arXiv
    2610.09027
    Published
    Layer
    Infrastructure layer
    Setting
    Models and APIs
    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
  12. Analysis and theoryarXiv 2610.09004

    研究：去除有害信息不足以证明开源权重模型的抗篡改能力

    论证发布时互信息不能认证开源权重抗微调恢复

    arXiv
    2610.09004
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    EleutherAI/deep-ignorance-e2e-strong-filter, EleutherAI/deep-ignorance-unfiltered, Google BERT-Tiny and others6
    Abstract作者称零信息仍可一步恢复，表观抗微调可来自参数化。

    作者研究开源权重模型的篡改抵抗：发布时的互信息是否足以保证微调恢复很慢。

    Paper analysisRead analysis
  14. InterpretabilityarXiv 2610.08887

    Moshi 全双工语音模型的情感引导遵循几何结构而非标签

    用均值差激活转向检验全双工语音模型的情感方向几何

    arXiv
    2610.08887
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    kyutai/moshika-pytorch-q8
    Abstract转向跟方向几何走而不跟标签走：sad 可单独对准，高唤醒情感共享方向。

    Moshi 上的情感激活转向跟方向之间的几何关系走，不跟被请求的标签走。

    Paper analysisRead analysis
  17. Evaluation and benchmarksarXiv 2610.07939

    CCDF dataset: Grok Imagine, Veo 3.1, Sora 2 forge surveillance footage, ten deepfake detectors basically cannot recognize it

    提出监控深伪基准 CCDF，在统一编码后评测检测器对伪造犯罪事故录像的识别

    arXiv
    2610.07939
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    C2P-CLIP, ForgeLens, D3-Im and others10

    Abstract提出消除格式捷径的监控深度伪造基准 CCDF，揭示现有检测器难以识别商用生成视频。

    Paper analysisRead analysis
  18. DefensesarXiv 2610.07774

    研究者提出读取路由 logits 的 MoE 视觉语言模型多模态安全检测器

    AI summary研究者提出一种轻量级路由 logits 安全检测器，在提示词预填充阶段读取 MoE 视觉语言模型的专家路由信号，在生成前判断多模态请求是否安全，且不修改模型参数或专家路由。

    arXiv
    2610.07774
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-VL-30B-A3B-Instruct, Kimi-VL-A3B-Instruct
    Abstract读出而非操纵 router logits，可在不改模型的情况下降低两模型的组合式安全错误。

    作者在两个指令微调 MoE VLM 上比较改模型与读路由。组合式风险按 HoliSafe 分成五类：图文各自可安全，合在一起才不安全，或不安全内容只在图像里。

    Paper analysisRead analysis
  19. AttacksarXiv 2610.07723

    Researchers propose answer-side backdoors: letting the model self-generate trigger words to bypass safety refusals.

    提出回答端后门，让模型自生成触发词以在干净有害请求下绕过拒答

    arXiv
    2610.07723
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Llama-2-7B-chat-hf, Mistral-7B-Instruct-v0.3, Gemma-7B-it and others4

    Abstract论文提出了多轮对话中回答端自生成触发词绕过安全拒绝的后门攻击，指出输入端防护存在盲区。

    Paper analysisRead analysis
  20. AttacksarXiv 2610.07654

    研究揭示策略蒸馏的安全后门风险

    测量带后门教师经 OPD 向干净学生迁移潜伏后门

    arXiv
    2610.07654
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Qwen2.5-1.5B-Instruct, Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct
    Abstract作者称 OPD 可在压低无触发词有害率时迁入后门。

    带后门的外部教师经 OPD，可在压低无触发词有害率的同时，把触发条件下的有害行为传给初始干净学生。。

    Paper analysisRead analysis
Load more