gongzhitaao / adversarial-classifier

It turns out that adversarial and clean data are not twins, not at all.

Adversarial Classifier

This repo contains the code to reproduce the experiment in our paper (https://arxiv.org/abs/1704.04960), specifically the code to generate the figures and tables.

Code Dependencies

1. Python 3.6+
2. Keras https://keras.io/
3. Tensorflow https://tensorflow.org
4. tensorflow-adversarial https://github.com/gongzhitaao/tensorflow-adversarial

Datasets

We used MNIST, CIFAR10 and SVHN in our experiment.

Document

The paper itself is written in Org mode (http://orgmode.org/) employing the ICML2017 LaTeX template, with slight modification, i.e., removing the conference information.

adversarial-classifier's Issues

May I use it?

Hello,sir.
May I need to train the binary classfier? Can not I use it to detect adversarial samples？

Little Doubt about Article

Dear Sir,

I just took my first step in scientific research focused on the algorithms of defense adversarial samples. Recently I read your paper 《Adversarial and Clean Data Are Not Twins》 .I think it's very interesting and your clear and compact declaration inspired me a lot. But I got some doubt about it.

In my opinion,I think beacause the binary classifier has very high false negative,it may tends to recognize them as clean samples.

Maybe I make something wrong for this.could you please deal with my doubt? I would be very appreciated if you can help me.

Thank you very much!

Could you provide the file named fgsm

Could you provide a file named fgsm,thank you

where is the from attacks.fgsm import fgsm

hello，sir：
the line:from attacks.fgsm import fgsm
I can't find this file for attacks.fgsm,can you help me?