研究者提出 Privacy-Leaking Watermarks,可在本地多模态模型中泄露对话隐私
提出 PLW,用微调把先前对话敏感触发编码进后续生成图像
- arXiv
- 2610.05453
- Published
- Setting
- Models and APIs
- Evaluated models
- OmniGen2, BAGEL-7B
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
AbstractPLW 把先前对话线索写成后续图像中的隐水印,检出随触发、模型和间隔变化。