shorya-sharma/game-theory-for-adversarial-attacks-and-defenses

This repo contains the code for the preprint 'Game Theory for Adversarial Attacks and Defenses' (https://arxiv.org/abs/2110.06166v2)

Game Theory For Adversarial Attacks And Defenses

Attacks

• All code (.ipynb files) related to attack experiments is in the attacks/ folder.
• Our FGSM and MI-FGSM implementations are in the [summarize/attack.py]
• Caveat:
  • We use Google Colab Pro to run the code. Some blocks are related to loading the gdrive folder.
  • There are some blocks related to the checkpoints of our pre-trained classifiers. If you are interested in getting these checkpoint data, please contact us.

Random Initialization

SAP

• renet18.py is the file for building network without SAP moudle.
• renet18SAP.py is the file for building network with SAP moudle.
• SAP-like-networks.ipynb is the file for generating multiple SAP-like networks.

Super-Resolution-Based Defense

• clean_images is the folder to store the original clean images.
• perturbed_images is the folder to store the adversarial attack images.
• denoised_images is the folder to store the wavelet denoised images.
• recoverd_images is the folder to store the super resolved images.
• attack.py is the file to generate attack samples.
• cifarresnet.py is the file for defense classification.
• components.py is the file for network components.
• edsr.py is the file for image super resolution using EDSR.
• denoising.py is the file for image wavelet denoising.

game-theory-for-adversarial-attacks-and-defenses's Issues

kindly provide the reference paper for the same.

kindly provide the reference paper for the same.