Skip to content

Latest commit

 

History

History
43 lines (22 loc) · 1.46 KB

README.md

File metadata and controls

43 lines (22 loc) · 1.46 KB

targeted-black-box-attack

Code of Targeted Black Box Adversarial Attack Method for Image Classification Models, published on IJCNN 2019.

1. Files

  • MNIST.py: train a model on MNIST dataset.
  • AdvMNIST.py: train an adversarial attack model on MNIST dataset.
  • CIFAR10.py: train a model on CIFAR10 dataset.
  • AdvCIFAR10.py: train an adversarial attack model on CIFAR10 dataset.
  • CIFAR100.py: train a model on CIFAR100 dataset.
  • AdvCIFAR100.py: train an adversarial attack model on CIFAR100 dataset.
  • CIFAR10.py: train a model on CIFAR10 dataset.
  • AdvCIFAR10.py: train a adversarial attack model on CIFAR10 dataset.
  • AdvMNIST_sk.py: train an adversarial model on MNIST dataset to attack Naive Bayes, Decision Tree, and Random Forest et al.
  • FashionMNIST.py: train a model on MNIST dataset.
  • AdvFashionMNIST.py: train an adversarial attack model on MNIST dataset.

2. Datasets

  • You can dump MNIST, FashionMNIST, CIFAR10, and CIFAR100 datasets into h5 files rather than download the files. (MNIST.h5, FashionMNIST.h5, CIFAR10.h5, CIFAR100.h5)

3. Test environment:

  • CPU: Intel E5-2430 X 2, GPU: NVIDIA GTX1070Ti, tensorflow 1.8-1.14

4. Notes:

  • You can change the net body to try different networks.
  • SimpleV1C, SimpleV3, and SimpleV7 in the code correspond to SmallNet, SimpleNet, and ConcatNet in the paper respectively.