Questions tagged [adversarial-machines]

18 questions
6
votes
1 answer

Adverserial images in TensorFlow

I am reading an article that explains how to trick neural networks into predicting any image you want. I am using the mnist dataset. The article provides a relatively detailed walk through but the person who wrote it is using Caffe. Anyways, my…
buydadip
  • 8,890
  • 22
  • 79
  • 154
2
votes
1 answer

Cannot run Carlini and Wagner Attack using foolbox on a tensorflow Model

I am using the latest version of foolbox (3.3.1), and my code simply load a RESNET-50 CNN, adds some layers for a transferred learning application, and loads the weights as follows. from numpy.core.records import array import tensorflow as tf from…
mad
  • 2,677
  • 8
  • 35
  • 78
2
votes
0 answers

What we cover with iterative Fast Gradient Sign Method (iterative-FGSM) that we can't cover with simple Fast Gradient Sign Method (FGSM)

FGSM and iterative-FGSM are two methods of generating adversarial samples. After visiting hundreds of sites and blogs to know that why in iterative-FGSM we perform multiple steps. And what we cover with multiple steps in (iterative-FGSM) that we…
2
votes
2 answers

How to update GAN Generator and Discriminator asynchronously in Tensorflow?

I want to develop a GAN with Tensorflow, with the Generator being an autoencoder and the Discriminator a Convolutional Neural Net with binary output. There is no problem to develop an autoencoder and the CNN, but my idea is to train 1 epoch for each…
razimbres
  • 4,715
  • 5
  • 23
  • 50
2
votes
1 answer

Get gradient value necessary to break an image

I've been experimenting with adversarial images and I read up on the fast gradient sign method from the following link https://arxiv.org/pdf/1412.6572.pdf... The instructions explain that the necessary gradient can be calculated using…
buydadip
  • 8,890
  • 22
  • 79
  • 154
1
vote
1 answer

Tensorflow - numpy gradient check doesnt work

I'm trying to estimate the gradient of a function by the finite difference method : finite difference method for estimating gradient TLDR: grad f(x) = [f(x+h)-f(x-h)]/(2h) for sufficiently small h. this is also used in the gradient check phase to…
m0ss
  • 334
  • 2
  • 4
  • 17
1
vote
1 answer

ModuleNotFoundError: No module named 'art.attacks'

I have already installed the module "art" yesterday, but when I try to run my code, it happens: Traceback (most recent call last): File "D:/Desktop/captcha/src1/adv_ex.py", line 10, in from art.attacks.evasion import…
CHI NEW
  • 33
  • 1
  • 5
1
vote
0 answers

Universal adversarial pertubation query/issue

I am Shashank V, a final year ECE student in India, an working on a side project on adversarial attacks on Image classification specifically for faces. I found this amazing paper title "Universal adversarial perturbations" in CVPR 2017 and wanted to…
1
vote
0 answers

how to feed raw audio as input to generative adversarial network?

I have followed tutorials to generate image using generative adversarial network but i have no idea on how to feed audio as input to network .Do i need to do some transformation on the audio data or can give directly to the network.
Deivapriya
  • 11
  • 2
0
votes
0 answers

How to program a Byzantine node?

I have a network simulator, and I need to program a Byzantine node class for that simulator. The purpose of these Byzantine nodes is to test the resiliency of an arbitrary networking protocol. Implementing random behaviour would be easy, but not…
Zaz
  • 46,476
  • 14
  • 84
  • 101
0
votes
0 answers

partially initialized module 'xgboost' has no attribute 'DMatrix' (most likely due to a circular import)

I am running into an issue using the xgboost classifer from adversarial robustness toolbox (ART). AttributeError: partially initialized module 'xgboost' has no attribute 'DMatrix' (most likely due to a circular import) Steps: clone the ART…
Evan Gertis
  • 1,796
  • 2
  • 25
  • 59
0
votes
2 answers

one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [640]] is at version 4;

I want to use pytorch DistributedDataParallel for adversarial training. The loss function is trades.The code can run in DataParallel mode. But in DistributedDataParallel mode, I got this error. When I change the loss to AT, it can run successfully.…
0
votes
1 answer

Question on ElasticNet algorithm implemented in Cleverhans

I'm trying to use the Elastic-Net algorithm implemented in Cleverhans to generate adversarial samples in a classification task. The main problem is that i'm trying to use it in a way to obtain an higher confidence at classification time on a target…
0
votes
0 answers

Can we train specific part of tensor with tebnsorflow?

I am trying to make an adversarial image for the inceptionV3 model with tensorflow. For that I use a specific loss on the pixel of my input image. This works well model_input_layer = model.layers[0].input model_output_layer =…
Samos
  • 111
  • 6
0
votes
1 answer

Extracting original image format after adversarial attack with Cleverhans

Suppose I load up the MNIST dataset with Cleverhans and attack an image with FGM. Any image I load via the Cleverhans MNIST dataset already has its pixel values constrained to [0, 1], and the same is true after I attack the image (suppose I clip the…
1
2