Fix _differential_evolution.py and JSMA attack pytest error

Harry24k / adversarial-attacks-pytorch

PyTorch implementation of adversarial attacks [torchattacks].

https://adversarial-attacks-pytorch.readthedocs.io/en/latest/index.html

MIT License

1.79k stars 337 forks source link

Fix _differential_evolution.py and JSMA attack pytest error #139

Closed rikonaka closed 1 year ago

rikonaka commented 1 year ago

PR Type and Checklist

What kind of change does this PR introduce?

[x] FIX BUG
- [x] I'm very sorry, I was out for a walk tonight then I suddenly realized that code I just submitted today had some problems with the division by zero in the _differential_evolution.py file. The judgment for this case should be a little more complicated.
- [x] Fix JSMA attack pytest error as check say.

Harry24k commented 1 year ago

You are always one step ahead of me! Thank you for your work!

Harry24k commented 1 year ago

During checking the attack success rate of EADs, I find that the targeted versions of EADs have difficulty in making adversarial examples: https://github.com/Harry24k/adversarial-attacks-pytorch/blob/master/code_coverage/on_server.ipynb. SPSA also shows 0% attack success rate, but I guess this comes from SPSA being a black-box attack. Could you check the issue?

rikonaka commented 1 year ago

During checking the attack success rate of EADs, I find that the targeted versions of EADs have difficulty in making adversarial examples: https://github.com/Harry24k/adversarial-attacks-pytorch/blob/master/code_coverage/on_server.ipynb. SPSA also shows 0% attack success rate, but I guess this comes from SPSA being a black-box attack. Could you check the issue?

OK, I will check it as soon as possible. 🤯