Research map / Image generation / Autoregressive and latent generation
DiffUE: Enhancing Utility-Unlearnability Trade-off of Unlearnable Examples via Diffusion Autoencoders
Syed Irfan Ali Meerza, Oktay Ozturk, Amir Sadovnik, Jian Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| inject noise in semantic space instead of pixel space (vs SEM) | DiffUE · CIFAR-10, adversarial training rho_t=4/255 | test accuracy (lower is better): 46.85 -> 17.11 | Improved |
| semantic noise vs pixel noise under adversarial training | DiffUE · ImageNet Subset, rho_t=4/255 | test accuracy (lower is better): 38.57 -> 17.20 | Improved |
| semantic noise under mean filtering | DiffUE · CIFAR-10 | test accuracy (lower is better): 28.55 -> 13.32 | Improved |
| semantic noise against UEraser | DiffUE · CIFAR-10 | test accuracy (lower is better): 81.56 -> 14.19 | Improved |
| semantic noise against LUE diffusion purification | DiffUE · CIFAR-10 | test accuracy (lower is better): 89.29 -> 33.97 | Improved |
| semantic noise under JPEG compression | DiffUE · CIFAR-10 | test accuracy (lower is better): 45.22 -> 17.22 | Improved |
| semantic noise vs pixel noise without any defense | DiffUE · ImageNet, no compression/attack | test accuracy (lower is better): 1.89 -> 2.11 | No clear effect |
| semantic noise transferred from ResNet-18 to VGG-16 | DiffUE · CIFAR-10 | test accuracy (lower is better): 54.55 -> 25.33 | Improved |
| semantic noise improves perceptual quality (FID) | DiffUE · CelebA-HQ | FID: 10.145 -> 5.355 | Improved |
| semantic noise improves image naturalness in user study | DiffUE · User study | naturalness: 3.11 -> 4.65 | Improved |
| larger semantic noise radius for stronger unlearnability | DiffUE · semantic radius selection | unlearnability vs semantics | Mixed |
Closest papers by meaning
- The Illusion of Unlearning: The Unstable Nature of Machine Unlearning in Text-to-Image Diffusion Models (CVPR 2025 · similarity 0.89)
- UnlearnCanvas: Stylized Image Dataset for Enhanced Machine Unlearning Evaluation in Diffusion Models (NeurIPS 2024 · similarity 0.88)
- Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models (NeurIPS 2026 · similarity 0.87)
- SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders (ICML 2025 · similarity 0.87)
- Diff-Plugin: Revitalizing Details for Diffusion-based Low-level Tasks (CVPR 2024 · similarity 0.86)
- Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
