最近一年的 AI 安全论文
本站共收录 68 篇- MMAligner：通过表示校准保护多模态大语言模型
MMAligner: Safeguarding Multimodal Large Language Models through Representation Calibration
- 对齐是局部的：用户侧说服下 GUI 智能体的配对诊断
Alignment Is Local: A Paired Diagnostic for GUI Agents under User-Side Persuasion
- 论成功与简洁：重新审视可迁移的视觉语言攻击流程
On Success and Simplicity: A Second Look at Transferable Vision-Language Attack Pipeline
- SeClaw 发布面向自主 Agent 的规格驱动安全任务合成与评测框架
SeClaw: Spec-Driven Security Task Synthesis for Evaluating Autonomous Agents
- CMTA：利用跨模态时序伪影实现可泛化的 AI 生成视频检测
CMTA: Leveraging Cross-Modal Temporal Artifacts for Generalizable AI-Generated Video Detection
- MGTEVAL：系统性评测机器生成文本检测器的交互式平台
MGTEVAL: An Interactive Platform for Systemtic Evaluation of Machine-Generated Text Detectors
- 当安全概念变得不安全：文生图模型中的多概念组合性漏洞
When Safe Concepts Become Unsafe: Multi-Concept Compositional Vulnerabilities in Text-to-Image Models
- 当理解成为风险：新兴图像生成范式 MLLM 的真实性与安全风险
When Understanding Becomes a Risk: Authenticity and Safety Risks in the Emerging Image Generation Paradigm
- Shell 中的假朋友：揭示大语言模型的表情符号语义混淆
False Friends in the Shell: Unveiling the Emoticon Semantic Confusion in Large Language Models
- 少关注虚词，让视觉语言模型免费获得鲁棒性
Pay Less Attention to Function Words for Free Robustness of Vision-Language Models
- GEO-Detective：用 LLM 智能体揭示图像中的位置隐私风险
GEO-Detective: Unveiling Location Privacy Risks in Images with LLM Agents
- 针对自动驾驶交通标志识别系统的 T2I 物理世界外观攻击
T2I-Based Physical-World Appearance Attack against Traffic Sign Recognition Systems in Autonomous Driving
- 基于 FPN-Transformer 的多模态深度伪造检测与定位
Multi-modal Deepfake Detection and Localization with FPN-Transformer
- PATN：面向移动传感器数据的实时隐私保护预测对抗变换网络
Privacy on the Fly: A Predictive Adversarial Transformation Network for Mobile Sensor Data
署名以 dblp 和本人的 Google Scholar 为准；只列本站收录的 AI 安全论文。