https://github.com/causallearning/hsja

https://github.com/causallearning/hsja

Science Score: 36.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
  • Academic publication links
    Links to: arxiv.org
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (10.0%) to scientific vocabulary
Last synced: 10 months ago · JSON representation

Repository

Basic Info
  • Host: GitHub
  • Owner: CausalLearning
  • Language: Python
  • Default Branch: master
  • Size: 4.01 MB
Statistics
  • Stars: 111
  • Watchers: 0
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created almost 2 years ago · Last pushed almost 2 years ago
Metadata Files
Readme

README.md

HopSkipJumpAttack

(previously named as Boundary Attack++) Code for HopSkipJumpAttack: A Query-Efficient Decision-Based Adversarial Attack by Jianbo Chen, Michael I. Jordan, Martin J. Wainwright.

Clarification

If you are using the code to compare HSJA with other algorithms, you will need to set your own counter for the number of model evaluations. "maxnumevals" in hsja.py is the maximum number of evaluations for estimating gradient (for each iteration), not the total number of model evaluations. To increase the total number of model evaluations, set a larger num_iterations.

Dependencies

The code for HopSkipJumpAttack runs with Python and requires Tensorflow of version 1.2.1 or higher. Please pip install the following packages: - numpy - tensorflow - keras - scipy

Running in Docker, MacOS or Ubuntu

We provide as an example the source code to run HopSkipJumpAttack on a ResNet trained on CIFAR-10. Run the following commands in shell:

```shell

Omit if already git cloned.

git clone https://github.com/Jianbo-Lab/HSJA cd HSJA

#########################################

Carry out L2 based untargeted attack on 5 samples.

python main.py --constraint l2 --attacktype untargeted --numsamples 5

Carry out L2 based targeted attack on 5 samples.

python main.py --constraint l2 --attacktype targeted --numsamples 5

Carry out Linf based untargeted attack on 5 samples.

python main.py --constraint linf --attacktype untargeted --numsamples 5

Carry out Linf based targeted attack on 5 samples.

python main.py --constraint linf --attacktype targeted --numsamples 5

Results are stored in cifar10resnet/figs/.

For each image, the left-hand side is the original example and

the right-hand side is its perturbed version.

```

See main.py and hsja.py for details.

Citation

If you use this code for your research, please cite our paper: @article{chen2019boundary, title={HopSkipJumpAttack: A Query-Efficient Decision-Based Adversarial Attack}, author={Chen, Jianbo and Jordan, Michael I. and Wainwright, Martin J.}, journal={arXiv preprint arXiv:1904.02144}, year={2019} }

Owner

  • Name: CausalLearning
  • Login: CausalLearning
  • Kind: organization

GitHub Events

Total
  • Watch event: 66
Last Year
  • Watch event: 66