https://github.com/shreyansh26/convnext-adversarial-examples

Generating Adversarial examples for ConvNeXt
https://github.com/shreyansh26/convnext-adversarial-examples

adversarial-attacks adversarial-machine-learning convnext deeplearning image-classification

Last synced: 7 months ago
JSON representation

Generating Adversarial examples for ConvNeXt

• Host: GitHub
• URL: https://github.com/shreyansh26/convnext-adversarial-examples
• Owner: shreyansh26
• Created: 2022-01-16T17:53:52.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-01-16T18:57:26.000Z (over 3 years ago)
• Last Synced: 2025-01-14T02:14:23.172Z (9 months ago)
• Topics: adversarial-attacks, adversarial-machine-learning, convnext, deeplearning, image-classification
• Language: Jupyter Notebook
• Homepage:
• Size: 6.85 MB
• Stars: 4
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Adversarial examples generation for ConvNeXt

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1c7EiO59cmVbxdCFZRDn82I-gBi6aTPDN?usp=sharing)

**This project is a Pytorch implementation of [@stanislavfort's project](https://twitter.com/stanislavfort/status/1481263565998805002?s=20).**

The notebook looks at generating adversarial images to "fool" the ConvNeXt model's image classification capabilities. [ConvNeXt](https://arxiv.org/abs/2201.03545) came out earlier this year from Meta AI.

The FGSM (Fast Gradient Sign Method) is a great algorithm to attack models in a *white-box* fashion with the goal of misclassification. Noise is added to the input image (not randomly) but in a manner such that the direction is the same as the gradient of the cost function with respect to the data.

Since this notebook is just the implementation - If you want to know more about FGSM, you may refer these - 

1. https://www.tensorflow.org/tutorials/generative/adversarial_fgsm

2. https://pytorch.org/tutorials/beginner/fgsm_tutorial.html

3. https://arxiv.org/abs/1412.6572

The following figure summarizes the goal of this notebook -