Summary of Boosting Imperceptibility Of Stable Diffusion-based Adversarial Examples Generation with Momentum, by Nashrah Haque et al.
Boosting Imperceptibility of Stable Diffusion-based Adversarial Examples Generation with Momentum
by Nashrah Haque, Xiang Li, Zhehui Chen, Yanzhao Wu, Lei Yu, Arun Iyengar, Wenqi Wei
First submitted to arxiv on: 17 Oct 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary We propose Stable Diffusion-based Momentum Integrated Adversarial Examples (SD-MIAE), a novel framework for generating visually imperceptible yet misclassifying adversarial examples. SD-MIAE leverages the text-to-image generation capabilities of the Stable Diffusion model by manipulating token embeddings to produce natural-looking images that can effectively mislead neural network classifiers. The framework consists of two phases: an initial adversarial optimization phase and a momentum-based optimization phase, which refines perturbations across iterations. Experimental results demonstrate SD-MIAE achieves a high misclassification rate of 79%, improving by 35% over the state-of-the-art method while preserving imperceptibility and semantic similarity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary We came up with a new way to trick neural networks called SD-MIAE (Stable Diffusion-based Momentum Integrated Adversarial Examples). It’s like generating fake images that can fool machines, but still look realistic. Our method uses a special model that can turn text into images and manipulates the image-making process to create these fake images. These fake images are designed to be hard for machines to classify correctly. We tested our method and it worked really well, being able to trick 79% of machine classifiers. |
Keywords
» Artificial intelligence » Diffusion » Diffusion model » Image generation » Neural network » Optimization » Token
