Pytorch | 利用MI-FGSM针对CIFAR10上的ResNet分类器进行对抗攻击
Pytorch | 利用MI-FGSM针对CIFAR10上的ResNet分类器进行对抗攻击
- CIFAR数据集
- MI-FGSM介绍
- 背景
- 算法原理
- MI-FGSM代码实现
- MI-FGSM算法实现
- 攻击效果
- 代码汇总
- mifgsm.py
- train.py
- advtest.py
之前已经针对CIFAR10训练了多种分类器:
Pytorch | 从零构建AlexNet对CIFAR10进行分类
Pytorch | 从零构建Vgg对CIFAR10进行分类
Pytorch | 从零构建GoogleNet对CIFAR10进行分类
Pytorch | 从零构建ResNet对CIFAR10进行分类
Pytorch | 从零构建MobileNet对CIFAR10进行分类
Pytorch | 从零构建EfficientNet对CIFAR10进行分类
Pytorch | 从零构建ParNet对CIFAR10进行分类
本篇文章我们使用Pytorch实现MI-FGSM对CIFAR10上的ResNet分类器进行攻击.
CIFAR数据集
CIFAR-10数据集是由加拿大高级研究所(CIFAR)收集整理的用于图像识别研究的常用数据集,基本信息如下:
- 数据规模:该数据集包含60,000张彩色图像,分为10个不同的类别,每个类别有6,000张图像。通常将其中50,000张作为训练集,用于模型的训练;10,000张作为测试集,用于评估模型的性能。
- 图像尺寸:所有图像的尺寸均为32×32像素,这相对较小的尺寸使得模型在处理该数据集时能够相对快速地进行训练和推理,但也增加了图像分类的难度。
- 类别内容:涵盖了飞机(plane)、汽车(car)、鸟(bird)、猫(cat)、鹿(deer)、狗(dog)、青蛙(frog)、马(horse)、船(ship)、卡车(truck)这10个不同的类别,这些类别都是现实世界中常见的物体,具有一定的代表性。
下面是一些示例样本:
MI-FGSM介绍
MI-FGSM(Momentum Iterative Fast Gradient Sign Method)是一种基于动量的迭代快速梯度符号法,是在FGSM(Fast Gradient Sign Method)基础上的改进,旨在生成更具攻击性和隐蔽性的对抗样本,以下是对其的详细介绍:
背景
- 在对抗攻击领域,FGSM是一种简单有效的攻击方法,但它仅进行一次梯度计算和更新,生成的对抗样本可能不够强大。为了进一步提高攻击效果,研究人员提出了迭代攻击的方法,如I-FGSM(Iterative FGSM),通过多次迭代来逐步调整对抗样本。MI-FGSM在I-FGSM的基础上引入动量项,使得攻击能够更好地利用历史梯度信息,加速收敛并提高攻击成功率。
算法原理
- 初始化:与FGSM类似,首先需要一个预训练的模型、损失函数、原始图像和对应的真实标签,以及攻击步长 ϵ \epsilon ϵ 、迭代次数 T T T和动量因子 μ \mu μ等参数。
- 迭代更新:在每次迭代中,计算当前对抗样本相对于模型输出的损失梯度,并将其与上一次迭代的动量项相加,得到更新后的梯度方向。然后,根据更新后的梯度方向和攻击步长,对对抗样本进行更新。具体计算公式如下:
g t + 1 = μ ⋅ g t + ∇ x J ( x t a d v , y ) ∥ ∇ x J ( x t a d v , y ) ∥ 1 g_{t+1}=\mu \cdot g_{t}+\frac{\nabla_{x} J\left(x_{t}^{adv}, y\right)}{\left\|\nabla_{x} J\left(x_{t}^{adv}, y\right)\right\|_{1}} gt+1=μ⋅gt+∥∇xJ(xtadv,y)∥1∇xJ(xtadv,y)
x t + 1 a d v = x t a d v + ϵ ⋅ sign ( g t + 1 ) x_{t+1}^{adv}=x_{t}^{adv}+\epsilon \cdot \text{sign}\left(g_{t+1}\right) xt+1adv=xtadv+ϵ⋅sign(gt+1)
其中, g t g_{t} gt 是第 t t t次迭代的动量项, x t a d v x_{t}^{adv} xtadv是第 t t t次迭代得到的对抗样本, J J J是损失函数, ∇ x J ( x t a d v , y ) \nabla_{x} J\left(x_{t}^{adv}, y\right) ∇xJ(xtadv,y) 是损失函数关于对抗样本的梯度, sign \text{sign} sign 表示符号函数。 - 投影操作:为了确保对抗样本在合理的范围内,通常还需要进行投影操作,将其像素值限制在有效区间内,如 [ 0 , 1 ] [0, 1] [0,1] 或 [ − 1 , 1 ] [-1, 1] [−1,1] 。
MI-FGSM代码实现
MI-FGSM算法实现
import torch
import torch.nn as nndef MI_FGSM(model, criterion, original_images, labels, epsilon, alpha=0.001, num_iterations=10, decay=1):"""MI-FGSM (Momentum Iterative Fast Gradient Sign Method) 参数:- model: 要攻击的模型- criterion: 损失函数- original_images: 原始图像- labels: 原始图像的标签- epsilon: 最大扰动幅度- alpha: 每次迭代的步长- num_iterations: 迭代次数- decay: 动量衰减因子"""# 复制原始图像作为初始的对抗样本perturbed_image = original_images.clone().detach().requires_grad_(True)momentum = torch.zeros_like(original_images).detach().to(original_images.device)for _ in range(num_iterations):outputs = model(perturbed_image)loss = criterion(outputs, labels)model.zero_grad()loss.backward()data_grad = perturbed_image.grad.data# 更新动量 (batch_size, channels, height, width)momentum = decay * momentum + data_grad / torch.sum(torch.abs(data_grad), dim=(1, 2, 3), keepdim=True)# 计算带动量的符号梯度sign_data_grad = momentum.sign()# 更新对抗样本perturbed_image = perturbed_image + alpha * sign_data_gradperturbed_image = torch.clamp(perturbed_image, original_images - epsilon, original_images + epsilon).detach().requires_grad_(True)return perturbed_image
攻击效果
代码汇总
mifgsm.py
import torch
import torch.nn as nndef MI_FGSM(model, criterion, original_images, labels, epsilon, alpha=0.001, num_iterations=10, decay=1):"""MI-FGSM (Momentum Iterative Fast Gradient Sign Method) 参数:- model: 要攻击的模型- criterion: 损失函数- original_images: 原始图像- labels: 原始图像的标签- epsilon: 最大扰动幅度- alpha: 每次迭代的步长- num_iterations: 迭代次数- decay: 动量衰减因子"""# 复制原始图像作为初始的对抗样本perturbed_image = original_images.clone().detach().requires_grad_(True)momentum = torch.zeros_like(original_images).detach().to(original_images.device)for _ in range(num_iterations):outputs = model(perturbed_image)loss = criterion(outputs, labels)model.zero_grad()loss.backward()data_grad = perturbed_image.grad.data# 更新动量 (batch_size, channels, height, width)momentum = decay * momentum + data_grad / torch.sum(torch.abs(data_grad), dim=(1, 2, 3), keepdim=True)# 计算带动量的符号梯度sign_data_grad = momentum.sign()# 更新对抗样本perturbed_image = perturbed_image + alpha * sign_data_gradperturbed_image = torch.clamp(perturbed_image, original_images - epsilon, original_images + epsilon).detach().requires_grad_(True)return perturbed_image
train.py
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
from models import ResNet18# 数据预处理
transform_train = transforms.Compose([transforms.RandomCrop(32, padding=4),transforms.RandomHorizontalFlip(),transforms.ToTensor(),transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
])transform_test = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
])# 加载Cifar10训练集和测试集
trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=False, transform=transform_train)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=128, shuffle=True, num_workers=2)testset = torchvision.datasets.CIFAR10(root='./data', train=False, download=False, transform=transform_test)
testloader = torch.utils.data.DataLoader(testset, batch_size=100, shuffle=False, num_workers=2)# 定义设备(GPU或CPU)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")# 初始化模型
model = ResNet18(num_classes=10)
model.to(device)# 定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.01)if __name__ == "__main__":# 训练模型for epoch in range(10): # 可以根据实际情况调整训练轮数running_loss = 0.0for i, data in enumerate(trainloader, 0):inputs, labels = data[0].to(device), data[1].to(device)optimizer.zero_grad()outputs = model(inputs)loss = criterion(outputs, labels)loss.backward()optimizer.step()running_loss += loss.item()if i % 100 == 99:print(f'Epoch {epoch + 1}, Batch {i + 1}: Loss = {running_loss / 100}')running_loss = 0.0torch.save(model.state_dict(), f'weights/epoch_{epoch + 1}.pth')print('Finished Training')
advtest.py
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
from models import *
from attacks import *
import ssl
import os
from PIL import Image
import matplotlib.pyplot as pltssl._create_default_https_context = ssl._create_unverified_context# 定义数据预处理操作
transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.491, 0.482, 0.446), (0.247, 0.243, 0.261))])# 加载CIFAR10测试集
testset = torchvision.datasets.CIFAR10(root='./data', train=False,download=False, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=128,shuffle=False, num_workers=2)# 定义设备(GPU优先,若可用)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")model = ResNet18(num_classes=10).to(device)criterion = nn.CrossEntropyLoss()# 加载模型权重
weights_path = "weights/epoch_10.pth"
model.load_state_dict(torch.load(weights_path, map_location=device))if __name__ == "__main__":# 在测试集上进行FGSM攻击并评估准确率model.eval() # 设置为评估模式correct = 0total = 0epsilon = 0.01 # 可以调整扰动强度for data in testloader:original_images, labels = data[0].to(device), data[1].to(device)original_images.requires_grad = Trueattack_name = 'MI-FGSM'if attack_name == 'FGSM':perturbed_images = FGSM(model, criterion, original_images, labels, epsilon)elif attack_name == 'BIM':perturbed_images = BIM(model, criterion, original_images, labels, epsilon)elif attack_name == 'MI-FGSM':perturbed_images = MI_FGSM(model, criterion, original_images, labels, epsilon)perturbed_outputs = model(perturbed_images)_, predicted = torch.max(perturbed_outputs.data, 1)total += labels.size(0)correct += (predicted == labels).sum().item()accuracy = 100 * correct / total# Attack Success RateASR = 100 - accuracyprint(f'Load ResNet Model Weight from {weights_path}')print(f'epsilon: {epsilon}')print(f'ASR of {attack_name} : {ASR}%')