adversecleaner

These tools do not work in the real world for several reasons: first, people who train LoRA or models curate their datasets, removing adversarial noise is trivial: https://github.com/lllyasviel/AdverseCleaner ; second, if they are trying to defend themselves against training AI models, then it will probably do almost nothing at all, models trained from scratch will learn the distribution, even one where the samples have adversarial noise (which can only attack a frozen model, like a VAE), SDXL has a new VAE, so the VAE will just be more robust if there is adversarial noise on the images, because it will learn to ignore it, many other models do not have a VAE to begin with. Also, resizing the image (as is almost always done when training the model) will probably destroy most if not all of the adversarial noise.

The author of https://haveibeentrained.com/ made an article about it: https://spawning.substack.com/p/we-tested-glaze-art-cloaking What surprises me most is that the paper does not consider the most obvious case: resizing of images before training; people usually train an SD model on resolutions of 512/768, so the noise is destroyed to a large extent even without realizing it. Why resizing the images is so effective is shown in this article about adversarial attacks: https://towardsdatascience.com/know-your-enemy-7f7c5038bdf3, the model after being trained with adversarial training learns independently to rescale the images as an adversarial defence. They also not consider in the Countermeasures section the fact that people can use diffusion models for what they are trained to do, denoising the images, if you put adversarial noise on it y