黑白盒 对抗扰动攻击方法 现成调用库 Attacks Toolbox

Attacks Toolbox

1. Adversarial Robustness Toolbox：
   1. pip install adversarial-robustness-toolbox
   2. such: from art.attacks.evasion import DeepFool
      1. art.attacks.evasion — Adversarial Robustness Toolbox 1.14.1 documentation
2. Torchattacks
   1. pip install torchattacks

   2. from torchattacks import FGSM

   3. eps=  */255
      1. torchattacks — torchattacks v3.4.0 documentation
3. CleverHans:
   1. pip install cleverhans
   2. cleverhans.torch.attacks
      1. 1. GitHub - cleverhans-lab/cleverhans: An adversarial example library for constructing attacks, building defenses, and benchmarking both
4. Foolbox:
   1. Welcome to Foolbox Native — Foolbox 3.3.3 documentation 

    5.  advertorch

         User Guide：Welcome to Advertorch — advertorch documentation

              1. advertorch.attacks — advertorch documentation

                AdverTorch: A toolbox for adversarial robustness research - Borealis AI

        例子：

现在终端

pip install advertorch

代码： 

from advertorch.attacks import LinfPGDAttack

# 定义PGDAttack，使用模型 model 生成对抗样本
adversary = LinfPGDAttack(predict= model, eps=0.2, targeted=True)

adv_inputs = adversary.perturb(data, target)    # 创建攻击