Skip to content

Implemention of the paper "Towards Deep Learning Models Resistant to Adversarial Attacks" using PaddlePaddle

Notifications You must be signed in to change notification settings

Cadden/paddle-adver

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Implemention of the paper "Towards Deep Learning Models Resistant to Adversarial Attacks" using PaddlePaddle

  1. 复现论文 “Towards Deep Learning Models Resistant to Adversarial Attacks”

  2. 运行方式: 使用python train.py命令进行模型训练
    使用python eval.py命令进行测试(该步骤可以直接运行,预训练的参数已经保存在环境中)

  3. 在Mnist数据集上复现结果如下: PGD-steps100-restarts20-sourceA: 90.89% (论文中结果89.3%), PGD-steps100-restarts20-sourceA: 97.53% (论文中结果95.7%),
    PGD-steps40-restarts1-sourceB: 98.47% (论文中结果96.4%)

  4. nat:表示原始图像,avd表示生成的对抗样本图像

About

Implemention of the paper "Towards Deep Learning Models Resistant to Adversarial Attacks" using PaddlePaddle

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages