arXivSub Start free trial

ICCV 2025 Papers — Page 28

IEEE/CVF International Conference on Computer Vision · 2701 papers

ZIUM: Zero-Shot Intent-Aware Adversarial Attack on Unlearned Models

Hyun Jun Yook (Chung-Ang University), Youn Kyu Lee (Chung-Ang University)

GenerationAdversarial AttackDiffusion modelImage

🎯 What it does: A zero-shot, intent-aware adversarial attack method for unlearned models, ZIUM, is proposed, which can generate images containing forgotten concepts based on user intent prompts without additional optimization.