-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlp_pretrained_attack_func.py
172 lines (142 loc) · 7.26 KB
/
lp_pretrained_attack_func.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.tensorboard import SummaryWriter
import numpy as np
import os
import random
from generator_mask import Generator
from discriminator import Discriminator
from pixel_valuation import PVRL
models_path = "./models/lp_sep/"
class PVRL_Attack:
def __init__(self,
device,
model,
generator,
discriminator,
model_num_labels,
box_min,
box_max):
self.device = device
self.model_num_labels = model_num_labels
self.model = model
self.box_min = box_min
self.box_max = box_max
self.netG = generator
self.netDisc = discriminator
self.netPv = PVRL().to(device)
# initialize optimizers
self.optimizer_G = torch.optim.Adam(self.netG.parameters(),
lr=1e-3)
self.optimizer_D = torch.optim.Adam(self.netDisc.parameters(),
lr=1e-3)
self.optimizer_PV = torch.optim.Adam(self.netPv.parameters(),
lr=1e-2)
if not os.path.exists(models_path):
os.makedirs(models_path)
def train_batch(self, x, labels):
self.optimizer_PV.zero_grad()
# optimize D
for i in range(1):
# Mask
probs = self.netPv(x)
probs_cpu = probs.cpu().detach()
probs_np = probs_cpu.numpy()
mask = np.random.binomial(1, probs_np, probs_np.shape)
mask = torch.from_numpy(mask)
mask = mask.view(mask.size(0), 1, x.size(2), x.size(3))
mask = mask.type(torch.FloatTensor).to(self.device)
# Probs loss + Mask num pixels
epsilon = 1e-8
probs_arr = probs.view(probs.size(0), -1)
mask_arr = mask.view(mask.size(0), -1)
probs_loss = torch.sum(torch.log(probs_arr + epsilon) * mask_arr, dim=1) + torch.sum(torch.log(1 -probs_arr + epsilon) * (1 - mask_arr), dim=1)
mask_num_pixels = torch.sum(mask_arr, dim=1)
# Add mask as fourth channel + Generate perturbation
x_with_mask = torch.cat((x, mask), 1).to(self.device)
perturbation = self.netG(x_with_mask)
# add a clipping trick
adv_images = torch.clamp(perturbation, -0.3, 0.3) * mask + x
adv_images = torch.clamp(adv_images, self.box_min, self.box_max)
self.optimizer_D.zero_grad()
pred_real = self.netDisc(x)
loss_D_real = F.mse_loss(pred_real, torch.ones_like(pred_real, device=self.device))
loss_D_real.backward()
pred_fake = self.netDisc(adv_images.detach())
loss_D_fake = F.mse_loss(pred_fake, torch.zeros_like(pred_fake, device=self.device))
loss_D_fake.backward()
loss_D_GAN = loss_D_fake + loss_D_real
self.optimizer_D.step()
# optimize G
for i in range(1):
self.optimizer_G.zero_grad()
# cal G's loss in GAN
pred_fake = self.netDisc(adv_images)
loss_G_fake = F.mse_loss(pred_fake, torch.ones_like(pred_fake, device=self.device))
loss_G_fake.backward(retain_graph=True)
# calculate perturbation norm
loss_perturb = torch.mean(torch.norm(perturbation.view(perturbation.shape[0], -1) * mask.view(mask.shape[0], -1), 2, dim=1))
# loss_perturb = torch.max(loss_perturb - C, torch.zeros(1, device=self.device))
# cal adv loss
logits_model = self.model(adv_images)
probs_model = F.softmax(logits_model, dim=1)
onehot_labels = torch.eye(self.model_num_labels, device=self.device)[labels]
# C&W loss function
real = torch.sum(onehot_labels * probs_model, dim=1)
other, _ = torch.max((1 - onehot_labels) * probs_model - onehot_labels * 10000, dim=1)
zeros = torch.zeros_like(other)
loss_adv_arr = torch.max(real - other, zeros)
loss_adv = torch.sum(loss_adv_arr)
# maximize cross_entropy loss
# loss_adv = -F.mse_loss(logits_model, onehot_labels)
# loss_adv = - F.cross_entropy(logits_model, labels)
adv_lambda = 10
pert_lambda = 1
loss_G = adv_lambda * loss_adv + pert_lambda * loss_perturb
loss_G.backward(retain_graph=True)
self.optimizer_G.step()
average_pixels = torch.mean(mask_num_pixels)
rl_loss = torch.mean(-probs_loss * loss_adv_arr + torch.log(torch.sum(mask_num_pixels)))
rl_loss.backward()
self.optimizer_PV.step()
return loss_D_GAN.item(), loss_G_fake.item(), loss_perturb.item(), loss_adv.item(), rl_loss.item(), average_pixels.item()
def train(self, train_dataloader, epochs):
writer = SummaryWriter(log_dir='./visualization/lp_rl/sep/', comment='Limited Pixel Attack using Reinforcement Learning')
for epoch in range(1, epochs+1):
loss_D_sum = 0
loss_G_fake_sum = 0
loss_perturb_sum = 0
loss_adv_sum = 0
loss_rl_sum = 0
for i, data in enumerate(train_dataloader, start=0):
images, labels = data
images, labels = images.to(self.device), labels.to(self.device)
loss_D_batch, loss_G_fake_batch, loss_perturb_batch, loss_adv_batch, loss_rl_batch, average_pixels = \
self.train_batch(images, labels)
loss_D_sum += loss_D_batch
loss_G_fake_sum += loss_G_fake_batch
loss_perturb_sum += loss_perturb_batch
loss_adv_sum += loss_adv_batch
loss_rl_sum += loss_rl_batch
num_batch = len(train_dataloader)
# print statistics
print("epoch %d:\nloss_D: %.5f, loss_G_fake: %.5f,\
\nloss_perturb: %.5f, loss_adv: %.5f, \nloss_rl: %.5f, \nAverage pixels: %.5f\n" %
(epoch, loss_D_sum/num_batch, loss_G_fake_sum/num_batch,
loss_perturb_sum/num_batch, loss_adv_sum/num_batch,
loss_rl_sum/num_batch, average_pixels))
writer.add_scalar('advgan_discriminator_loss', loss_D_sum/num_batch, epoch)
writer.add_scalar('advgan_generator_loss', loss_G_fake_sum/num_batch, epoch)
writer.add_scalar('advgan_perturbation_loss', loss_perturb_sum/num_batch, epoch)
writer.add_scalar('advgan_adversarial_loss', loss_adv_sum/num_batch, epoch)
writer.add_scalar('pixel_valuation_loss', loss_rl_sum/num_batch, epoch)
writer.add_scalar('average_num_pixels', average_pixels, epoch)
netG_file_name = models_path + 'netG_rl_epoch_' + str(epoch) + '.pth'
torch.save(self.netG.state_dict(), netG_file_name)
netDisc_file_name = models_path + 'netDisc_rl_epoch_' + str(epoch) + '.pth'
torch.save(self.netDisc.state_dict(), netDisc_file_name)
netPv_file_name = models_path + 'netPv_rl_epoch_' + str(epoch) + '.pth'
torch.save(self.netPv.state_dict(), netPv_file_name)
writer.close()