activation-attack-pytorch's Introduction

Activation-Attack-Pytorch

Apply targeted attacks on black box ViT with the activation attack method. The white box model is ResNet-18.

Based on CVPR 2019 Paper Feature Space Perturbations Yield More Transferable Adversarial Examples.

My additional expreriment proves that the transferablitiy is low between heterogeneous deep models.

All the models are finetuned on CIFAR-10 with the Adam optimizer with lr = 1e-5.

Pretrained models

Model Name	CIFAR-10 Test Accuracy	ImageNet-1K Pretrain
ResNet-18	0.9435	Yes
DenseNet-121	0.9571	Yes
ViT-B/16	0.9838	Yes

Results

Model	Error	uTr	tSuc	tTr
ResNet-18	96.39	/	75.48	/
DenseNet-121	80.16	82.27	9.39	10.65
ViT	22.33	22.78	2.5	4.2

activation-attack-pytorch's People

Contributors

Stargazers

Watchers

activation-attack-pytorch's Issues

What about tsuc and ttr on ImageNet (ILSVRC2012)?

I run the code and change the dataset from CIFAR-10 to ILSVRC2012 validation set. The success rate on white box is always 0. It is a bit strange.
Moreover, I think the calculation about 'utr' and 'ttr' is not correct. Assuming that we have two minibatch with size 32. In one minibatch, wb_error is 3 and 2 of them are successfully transferred. Then utrs in this minibatch is 66%. While in the other minibatch, wb_error is 32 but none of them is successfully transferred. Then utrs is 0%. According the code, utr will be 33% = (66%+0)/2. But actually I think it should be 2/(3+32), not 33%.

qwq2000 / activation-attack-pytorch Goto Github PK

activation-attack-pytorch's Introduction

Activation-Attack-Pytorch

Pretrained models

Results

activation-attack-pytorch's People

Contributors

Stargazers

Watchers

Forkers

activation-attack-pytorch's Issues

What about tsuc and ttr on ImageNet (ILSVRC2012)?

可能的问题

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent