Skip to content

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

Oct 12 · Monday

All papers

Found 7 papers
Filter3

Attack type

Jailbreaks3

Defense type

Model hardening5Inference-time intervention6Detection and filtering14Instruction–data separation2Watermarking and provenance4

Impact

Output manipulation3

System type

Models and APIs6

Attacker access

White-box3

Attack surface / entry point

User input45Web pages and interfaces22Documents and messages13Tool responses28Tool and skill descriptions12Memory writes4Inter-agent messages4Images, audio and video7Training data21Weights and gradients12Distribution channels6Inference services and hardware3

Layer

Training and data layer3

System component

Vision4Multimodal6
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 265papers have no tags yet and are not included in filtered results.
Defense type：Model hardeningDefense type：Instruction–data separationAttack surface / entry point：Images, audio and videoClear all

Another 265papers do not have category tags yet and are not included in filtered results.

  1. Evaluation and benchmarksarXiv 2609.33481

    IDUnlearn-Bench：面向视觉语言模型的个体级遗忘基准

    提出 IDUnlearn-Bench，评测 VLM 个体级遗忘后的身份残留

    arXiv
    2609.33481
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-VL-2B, Qwen3-VL-4B, Qwen3-VL-8B and others6
    AbstractIDUnlearn-Bench 用四族探针显示：压低直接属性访问后，身份仍常能被反向查出、绑定或重构。

    IDUnlearn-Bench 评测 VLM 是否忘掉一个可跨模态识别的人，而不只是答不出某条监督事实。

    Paper analysisRead analysis
  4. DefensesarXiv 2609.23596

    StyleAT：用对抗训练防御人脸识别的语义攻击

    提出 BoundStyle 与 StyleAT，对抗训练防御人脸识别语义攻击

    arXiv
    2609.23596
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Attacker
    White-box, Black-box
    Evaluated models
    MobileFaceNet (MobileFace), ResNet (ResNet-152, IR-SE), RepVGG and others7
    AbstractBoundStyle 用 StyleGAN3 有界潜编辑快速攻击 FR。

    StyleAT 用可调的 StyleGAN3 潜空间攻击 BoundStyle 对人脸识别做对抗训练，以抵抗一般语义编辑，并在评测中面对训练时未见的 DiffPrivate。

    Paper analysisRead analysis
  5. DefensesarXiv 2609.10613

    后验重加权框架解释并缓解多模态大模型上下文越狱

    用后验重加权解释多模态上下文越狱，并提出风险门控的良性示例注入防御

    arXiv
    2609.10613
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-VL-2B, Qwen3-VL-4B, Qwen3-VL-8B and others8
    Abstract后验重加权解释上下文越狱的缩放，并支持风险门控的良性示例防御。

    Zhang 等把多模态 ICL 越狱写成对齐模型在安全与有害隐模式之间的推理时后验漂移，并据此做风险门控防御。

    Paper analysisRead analysis
  6. DefensesarXiv 2602.12461

    Semantic-aware adversarial fine-tuning for CLIP

    提出 SAFT，用语义描述集成构造对抗样本并微调图像编码器

    arXiv
    2602.12461
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    ViT-B/32, CLIP-L/14
    AbstractSAFT 用精炼类描述的集成对抗样本微调 CLIP 图像编码器，以改善零样本准确率与鲁棒性的权衡。

    SAFT 是对 CLIP 图像编码器做对抗微调的方法，要改善零样本分类上的准确率–鲁棒性权衡。

    Paper analysisRead analysis
  7. DefensesarXiv 2602.10915

    Blind Gods and Broken Screens: Building a Secure, Intent-Centric Mobile Agent Operating System

    提出 Aura，用可信内核中介移动 Agent 的身份、感知与动作控制

    arXiv
    2602.10915
    Published
    Layer
    Application layer
    Setting
    Browser and computer use
    Evaluated models
    Gemini-3-Flash
    AbstractAura 以 Agent Kernel 中介结构化交互。

    Aura 用内核中介的结构化智能体交互，替换以屏幕为接口的移动助手。。

    Paper analysisRead analysis
You have reached the end