Attax: adversarial attacks using JAX
Project description
Attax is python library with implementations of some common adversarial attacks using JAX. For now this is a prototype at best and I refer you my other project, Foolbox, that also supports JAX and provides a well-tested suite of adversarial attacks.
Installation
pip install attax
Example
import attax
from functools import partial
def predict(params, inputs):
# see https://github.com/google/jax
# ...
params = ... # model parameters
x = ... # input data
y = ... # labels
f = partial(predict, params)
x_adv = attax.pgd(f, x, y, epsilon=0.3)
print((f(x).argmax(axis=-1) == y).mean())
print((f(x_adv).argmax(axis=-1) == y).mean())
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
attax-0.0.2.tar.gz
(2.6 kB
view details)
File details
Details for the file attax-0.0.2.tar.gz
.
File metadata
- Download URL: attax-0.0.2.tar.gz
- Upload date:
- Size: 2.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/2.0.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.4.0 requests-toolbelt/0.9.1 tqdm/4.36.1 CPython/3.6.8
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 558fdbe91b9bd0029a59f449c31e0258c06dfc8540cf75678e8d47fd799c9b42 |
|
MD5 | 6640f134886fbd9aa5732381b6a76ab0 |
|
BLAKE2b-256 | 50d6799f34fc7d23d4a6fe406ccfb2d3c1868ad647bc47f646f23d3b50df9d37 |