search

Harry24k / adversarial-attacks-pytorch

PyTorch implementation of adversarial attacks [torchattacks]
https://adversarial-attacks-pytorch.readthedocs.io/en/latest/index.html
MIT License
1.86k stars 348 forks source link

Pixle attack implemented #60

Closed jaryP closed 2 years ago

jaryP commented 2 years ago

Hello Harry24k, first of all, I want to thank you for this useful and easy to use library.

I would like to add to it a new L0 clack-box attack that I, and others, have developed and published (you can see the original implementation here).

Looking forward to a reply, J

Harry24k commented 2 years ago

Hello, thank you for your contribution. Now I'm reading the paper and checking the performance of the attack.

Harry24k commented 2 years ago

Hi, I read the paper and it's quite interesting. I modified the following things:

By the way, I'm curious about the performance of 'swap' method, which is mentioned in the last paragraph on Page 5. Under the skim of 'no redundant information is injected', the swap method is better than the injection method.

Thank you for your awesome work.

p.s. I found the misspell on Page 5: p=(w_p, o_y, w_p, h_p) should be p=(o_x, o_y, w_p, h_p)