mlsanityData to 7 Oct 2026

Reversible Unlearnable Examples: Towards the Copyright Protection in Deep Learning Era

arXiv 2608.06211 ↗First posted 6 Aug 2026cs.CR cs.CVcited by 0

On the map: LLM safety and security → Watermarking generated content and models

Abstract (arXiv)

Significant advancements in deep learning have been made possible by the utilization of large datasets, underscoring the critical importance of copyright protection. Adding meticulously designed perturbations to examples, making them unlearnable has become a crucial approach for safeguarding data copyright. Existing methods for creating unlearnable examples overlook the risk of data leakage, which can threaten data ownership. Thus, copyright protection in deep learning faces two main threats: illegal model training and malicious data leakage. We investigate that these two threats cannot be solved by straightforwardly combining existing availability attacks and watermarking techniques as their negative interaction effects. Therefore, in this paper, we propose a novel copyright protection mechanism for the aforementioned security concerns. Considering that the prevention of unauthorized model training requires powerful generalizability of unlearnable perturbations, we generate perturbations to induce the model to learn uncorrelated features of input images. It works by minimizing the mutual information of the input and output of the model. On the other hand, to eliminate the side impact of unlearnable perturbations on the watermark extraction, we design a dual extraction strategy by using two distinct watermark extractors. Extensive experiments on the image datasets {ImageNet, CIFAR10, and Pets} show that our proposed method could provide comprehensive copyright protection to images. The code is available at {https://github.com/Yeah21/ReversibleUnlearnableExamples}.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Watermarking generated content and models, one of 12 directions in LLM safety and security. Builds on 39 of its 73 references that are in the map. First posted on arXiv 6 Aug 2026.

Builds on

39 of 73 references.

Closest ancestors · 1

Same direction, newest first

  1. 2021Copy, Right? A Testing Framework for Copyright Protection of Deep Learning Models10 Dec

Other references · 38

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  5. 2016Densely Connected Convolutional Networkscited by 6,691
  6. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,773
  7. 2017Squeeze-and-Excitation Networkscited by 4,763
  8. 2014Going Deeper with Convolutionscited by 4,231
  9. 2019Similarity of Neural Network Representations Revisitedcited by 1,695
  10. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,315
  11. 2016Delving into Transferable Adversarial Examples and Black-box Attackscited by 530
  12. 2018HiDDeN: Hiding Data With Deep Networkscited by 384
26 more
  1. 2018Insights on representational similarity in neural networks with canonical correlationcited by 318
  2. 2019StegaStamp: Invisible Hyperlinks in Physical Photographscited by 256
  3. 2020Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depthcited by 241
  4. 2021Revisiting Model Stitching to Compare Neural Representationscited by 203
  5. 2017Towards Poisoning of Deep Learning Algorithms with Back-gradient Optimizationcited by 188
  6. 2021Unlearnable Examples: Making Personal Data Unexploitablecited by 188
  7. 2020Witches' Brew: Industrial Scale Data Poisoning via Gradient Matchingcited by 156
  8. 2020Maximum-Entropy Adversarial Data Augmentation for Improved Generalization and Robustnesscited by 121
  9. 2021Adversarial Examples Make Strong Poisonscited by 115
  10. 2020MetaPoison: Practical General-purpose Clean-label Data Poisoningcited by 114
  11. 2021MBRS : Enhancing Robustness of DNN-based Watermarking by Mini-Batch of Real and Simulated JPEG Compressioncited by 108
  12. 2021Availability Attacks Create Shortcutscited by 60
  13. 2020Adv-watermark: A Novel Watermark Perturbation for Adversarial Examplescited by 56
  14. 2018ReDMark: Framework for Residual Diffusion Watermarking on Deep Networkscited by 55
  15. 2022Autoregressive Perturbations for Data Poisoningcited by 45
  16. 2022Transferable Unlearnable Examplescited by 41
  17. 2021Preventing Unauthorized Use of Proprietary Data: Poisoning for Secure Dataset Releasecited by 35
  18. 2022Unlearnable Clusters: Towards Label-agnostic Unlearnable Examplescited by 35
  19. 2022Robust Unlearnable Examples: Protecting Data Against Adversarial Learningcited by 34
  20. 2021Grounding Representation Similarity with Statistical Testingcited by 28
  21. 2023Learning the Unlearnable: Adversarial Augmentations Suppress Unlearnable Example Attackscited by 28
  22. 2020Adversarial Privacy-preserving Filtercited by 13
  23. 2024UnSeg: One Universal Unlearnable Example Generator is Enough against All Image Segmentationcited by 13
  24. 2022Learnability Lock: Authorized Learnability Control Through Adversarial Invertible Transformationscited by 7
  25. 2024ModelLock: Locking Your Model With a Spellcited by 6
  26. 2024A self-supervised CNN for image watermark removalcited by 3

Built on by

Not cited yet.