mlsanityData to 7 Oct 2026

Robustness, privacy and adversarial aspects

928 papers2 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: diffusion, adversarial, attacks, attack, concept

Key facts. Robustness, privacy and adversarial aspects: 928 papers on arXiv since 2020 (data through 7 Oct 2026). 216 of its 928 mapped papers (23%) cite “Erasing Concepts from Diffusion Models” (2023, arXiv 2303.07345). 208 of its 928 mapped papers (22%) cite “Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models” (2022, arXiv 2211.05105). 16 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

16 new papers, 17 Sep – 7 Oct; 2 this week.

  1. This week · 2
  2. On the Intrinsic Limited Robustness of Latent-Based Watermarking
  3. SPIN: Image Immunization Against Diffusion Editing via Single-Step Projection in Stochastic Neighborhoods also fits: Diffusion foundations and sampling
  4. Earlier · 14
  5. Your Unlearning Gives You Away: Identifying Erased Concepts in Diffusion Models
  6. No Concept Escapes the Audit: Auditing-Aware Unlearning for Verifiable Concept Erasure in Diffusion Models
  7. VisualErase: Dual-Branch Visual Trajectory Redirection for Robust Concept Erasure in Text-to-Image Diffusion Models
  8. Protecting Sensitive Data in Image Synthesis via PAC-Private Adaptation for Diffusion Models
  9. DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation also fits: Diffusion foundations and sampling
  10. Continual Concept Erasure in Diffusion Models by Suppressing Cross-Edit Interference
  11. Debias Anything: Fairness with Diversity without Supervision in Diffusion Models also fits: Image generation, text-to-image and editing
  12. Learning Normal Diffusion Dynamics for Backdoor Defense in Text-to-Image Models
All 16
  1. Weeding Out Bad Seeds: Initial-Noise-Robust Unlearning for Text-to-Image Diffusion Models also fits: Image generation, text-to-image and editing
  2. TANGO: Watermarking Masked Diffusion Language Models in Token Pairs
  3. You Can't Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning
  4. Concept Score Relearning: A Unified Cross-Architecture Attack on Concept Erasure
  5. FeatMark: Feature-level Watermark Protection against Mimicry Attacks with Diffusion Models
  6. LoRango: It Takes Two LoRAs to Unlock Hidden Behaviors in Diffusion Models

Foundations

Most cited by papers here

  1. Erasing Concepts from Diffusion Models2023cited by 216 here
  2. Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models2022cited by 208 here
  3. Unified Concept Editing in Diffusion Models2023cited by 184 here
  4. Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models2023cited by 149 here
  5. Ablating Concepts in Text-to-Image Diffusion Models2023cited by 147 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2020Denoising Diffusion Probabilistic Modelscited by 643 here
  2. 2020Denoising Diffusion Implicit Modelscited by 375 here
  3. 2021Improved Denoising Diffusion Probabilistic Modelscited by 114 here
  4. 2021Diffusion Models Beat GANs on Image Synthesiscited by 313 here
  5. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 643 here
  6. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 255 here
  7. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 266 here
  8. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 208 here
  9. 2023Erasing Concepts from Diffusion Modelscited by 216 here
  10. 2023Ablating Concepts in Text-to-Image Diffusion Modelscited by 147 here
  11. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Modelscited by 149 here
  12. 2023Unified Concept Editing in Diffusion Modelscited by 184 here
  13. 2023To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Nowcited by 116 here
  14. 2024MACE: Mass Concept Erasure in Diffusion Modelscited by 135 here
10 more it builds on, not in this chain
  1. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 245 here
  2. 2021GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Modelscited by 171 here
  3. 2022Classifier-Free Diffusion Guidancecited by 179 here
  4. 2022DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generationcited by 176 here
  5. 2022Red-Teaming the Stable Diffusion Safety Filtercited by 104 here
  6. 2023Extracting Training Data from Diffusion Modelscited by 137 here
  7. 2023Adding Conditional Control to Text-to-Image Diffusion Modelscited by 120 here
  8. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 159 here
  9. 2023Ring-A-Bell! How Reliable are Concept Removal Methods for Diffusion Models?cited by 131 here
  10. 2023One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applicationscited by 104 here

The 24 most cited of 1,782.

Browse

30 of 928, ranked 6–35

  1. 6Extracting Training Data from Diffusion Models2023137
  2. 7MACE: Mass Concept Erasure in Diffusion Models2024135
  3. 8Ring-A-Bell! How Reliable are Concept Removal Methods for Diffusion Models?2023131
  4. 9To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now2023116
  5. 10One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications2023104
  6. 11Red-Teaming the Stable Diffusion Safety Filter2022104
  7. 12Diffusion Models for Adversarial Purification2022103
  8. 13The Stable Signature: Rooting Watermarks in Latent Diffusion Models2023100
  9. 14SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation202398
  10. 15Prompting4Debugging: Red-Teaming Text-to-Image Diffusion Models by Finding Problematic Prompts202392
All 30
  1. 16MMA-Diffusion: MultiModal Attack on Diffusion Models202389
  2. 17Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models202488
  3. 18Raising the Cost of Malicious AI-Powered Image Editing202381
  4. 19Adversarial Example Does Good: Preventing Painting Imitation from Diffusion Models via Adversarial Examples202380
  5. 20Reliable and Efficient Concept Erasure of Text-to-Image Diffusion Models202477
  6. 21Receler: Reliable Concept Erasing of Text-to-Image Diffusion Models via Lightweight Erasers202372
  7. 22Are Diffusion Models Vulnerable to Membership Inference Attacks?202371
  8. 23Anti-DreamBooth: Protecting users from personalized text-to-image synthesis202368
  9. 24SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generation202467
  10. 25Tree-Ring Watermarks: Fingerprints for Diffusion Images that are Invisible and Robust202363
  11. 26Gaussian Shading: Provable Performance-Lossless Image Watermarking for Diffusion Models202459
  12. 27Text-to-Image Diffusion Models can be Easily Backdoored through Multimodal Data Poisoning202359
  13. 28SneakyPrompt: Jailbreaking Text-to-image Generative Models202359
  14. 29TrojDiff: Trojan Attacks on Diffusion Models with Diverse Targets202357
  15. 30A Recipe for Watermarking Diffusion Models202357
  16. 31Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-To-Image Models202357
  17. 32How to Backdoor Diffusion Models?202256
  18. 33VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models202348
  19. 34Mist: Towards Improved Adversarial Examples for Diffusion Models202346
  20. 35Guided Diffusion Model for Adversarial Purification202243