adversarial-library

This library contains various resources related to adversarial attacks implemented in PyTorch. It is aimed towards researchers looking for implementations of state-of-the-art attacks.

The code was written to maximize efficiency (e.g. by preferring low level functions from PyTorch) while retaining simplicity (e.g. by avoiding abstractions). As a consequence, most of the library, and especially the attacks, is implemented using pure functions (whenever possible).

While focused on attacks, this library also provides several utilities related to adversarial attacks: distances (SSIM, CIEDE2000, LPIPS), visdom callback, projections, losses and helper functions. Most notably the function run_attack from utils/attack_utils.py performs an attack on a model given the inputs and labels, with fixed batch size, and reports complexity related metrics (run-time and forward/backward propagations).

Dependencies

The goal of this library is to be up-to-date with newer versions of PyTorch so the dependencies are expected to be updated regularly (possibly resulting in breaking changes).

pytorch>=1.7.0
torchvision>=0.8.0
tqdm>=4.48.0
visdom>=0.1.8

Installation

You can either install using:

pip install git+https://github.com/jeromerony/adversarial-library

Or you can clone the repo and run:

python setup.py install

Alternatively, you can install (after cloning) the library in editable mode:

pip install -e .

Example

For an example on how to use this library, you can look at this repo: https://github.com/jeromerony/augmented_lagrangian_adversarial_attacks

Attacks

Currently the following attacks are implemented in the adv_lib.attacks module:

Name	Knowledge	Distance(s)	Link
Carlini and Wagner (C&W)	White-box	L₂, L_∞	https://arxiv.org/abs/1608.04644
Projected Gradient Descent (PGD)	White-box	L_∞	https://arxiv.org/abs/1706.06083
Decoupled Direction and Norm (DDN)	White-box	L₂	https://arxiv.org/abs/1811.09600
Trust Region (TR)	White-box	L₂, L_∞	https://arxiv.org/abs/1812.06371
Fast Adaptive Boundary (FAB)	White-box	L₁, L₂, L_∞	https://arxiv.org/abs/1907.02044
Perceptual Color distance Alternating Loss (PerC-AL)	White-box	CIEDE2000	https://arxiv.org/abs/1911.02466
Auto-PGD (APGD)	White-box	L₂, L_∞	https://arxiv.org/abs/2003.01690
Augmented Lagrangian Method for Adversarial (ALMA)	White-box	L₁, L₂, SSIM, CIEDE2000, LPIPS, ...	https://arxiv.org/abs/2011.11857
Voting Folded Gaussian Attack (VFGA)	White-box	L₀	https://arxiv.org/abs/2011.12423
Fast Minimum-Norm (FMN)	White-box	L₀, L₁, L₂, L_∞	https://arxiv.org/abs/2102.12827

Bold means that this repository contains the official implementation.

Distances

The following distances are available in the utils adv_lib.distances module:

Lp-norms
SSIM https://ece.uwaterloo.ca/~z70wang/research/ssim/
MS-SSIM https://ece.uwaterloo.ca/~z70wang/publications/msssim.html
CIEDE2000 color difference http://www2.ece.rochester.edu/~gsharma/ciede2000/ciede2000noteCRNA.pdf
LPIPS https://arxiv.org/abs/1801.03924

Contributions

Suggestions and contributions are welcome :)

Citation

If this library has been useful for your research, you can cite it as follows:

@misc{rony2020adversarial,
   title={Adversarial Library},
   author={Rony, J{\'e}r{\^o}me and {Ben Ayed}, Ismail},
   year={2020},
   url={https://github.com/jeromerony/adversarial-library}
}

shaodongheng / adversarial-library Goto Github PK