How do I implement alternate adversarial attacks in the tutorials

[ephi...@yahoo.com]

How do I implement an alternate adversarial attack (such as predicted gradient descent) in the tutorials cifar_tutorial_tf.py and/or mnist_tutorial_tf.py? Right now, these tutorials demonstrate the FGSM adversarial example. Thank you. Arnold

[Ian Goodfellow]

Copy paste the tutorial, import a different Attack subclass, and call it in the same places FGSM is called in the tutorial. You'll need to understand how the particular attack you want to use works and pass it appropriate arguments to configure it well.

On Mon, Oct 22, 2018 at 9:53 AM ephi5757 via cleverhans dev <cleverh...@googlegroups.com> wrote:

How do I implement an alternate adversarial attack (such as predicted gradient descent) in the tutorials cifar_tutorial_tf.py and/or mnist_tutorial_tf.py? Right now, these tutorials demonstrate the FGSM adversarial example. Thank you. Arnold

--
You received this message because you are subscribed to the Google Groups "cleverhans dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cleverhans-de...@googlegroups.com.
To post to this group, send email to cleverh...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cleverhans-dev/0fe779f7-d97c-4e1f-a8b6-808c6459ddb4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ephi...@yahoo.com]

Here is an example I executed of from the MNIST tensorflow tutorial for the input image and the FGSM adversarial attack with eps = 0.5

[ephi...@yahoo.com]

Terrific. Thank you for your help and rapid response

On Monday, October 22, 2018 at 12:57:28 PM UTC-4, Ian Goodfellow wrote:

Copy paste the tutorial, import a different Attack subclass, and call it in the same places FGSM is called in the tutorial. You'll need to understand how the particular attack you want to use works and pass it appropriate arguments to configure it well.

On Mon, Oct 22, 2018 at 9:53 AM ephi5757 via cleverhans dev <cleverhans-dev@googlegroups.com> wrote:

How do I implement an alternate adversarial attack (such as predicted gradient descent) in the tutorials cifar_tutorial_tf.py and/or mnist_tutorial_tf.py? Right now, these tutorials demonstrate the FGSM adversarial example. Thank you. Arnold

--
You received this message because you are subscribed to the Google Groups "cleverhans dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cleverhans-dev+unsubscribe@googlegroups.com.
To post to this group, send email to cleverhans-dev@googlegroups.com.