/
Februus.py
418 lines (351 loc) · 16.3 KB
/
Februus.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
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
import argparse
import logging
import os
import sys
sys.path.append('/')
sys.path.append(os.getcwd())
import time
import torch
import torch.nn as nn
import numpy as np
from utils.fix_random import fix_random
from utils.aggregate_block.dataset_and_transform_generate import get_transform, get_dataset_norm_stats
from utils.aggregate_block.model_trainer_generate import generate_cls_model
from utils.dataset.bd_dataset import prepro_cls_DatasetBD
from utils.save_load_attack import load_attack_result
import yaml
from pprint import pformat
from models.CompletionNetwork import CompletionNetwork
import cv2
import os
count = os.cpu_count()
torch.set_num_threads(min(count//2, 32))
def poisson_blend_old(input, output, mask):
"""
* inputs:
- input (torch.Tensor, required)
Input tensor of Completion Network.
- output (torch.Tensor, required)
Output tensor of Completion Network.
- mask (torch.Tensor, required)
Input mask tensor of Completion Network.
* returns:
Image tensor inpainted using poisson image editing method.
"""
num_samples = input.shape[0]
ret = []
mask = torch.cat((mask, mask, mask), dim=1) # convert to 3-channel format
# convert torch array to numpy array followed by
# converting 'channel first' format to 'channel last' format.
input_np = np.transpose(np.copy(input.cpu().numpy()), axes=(0, 2, 3, 1))
output_np = np.transpose(np.copy(output.cpu().numpy()), axes=(0, 2, 3, 1))
mask_np = np.transpose(np.copy(mask.cpu().numpy()), axes=(0, 2, 3, 1))
# apply poisson image editing method for each input/output image and mask.
for i in range(num_samples):
inpainted_np = newblend(input_np[i], output_np[i], mask_np[i])
inpainted = torch.from_numpy(np.transpose(inpainted_np, axes=(2, 0, 1)))
inpainted = torch.unsqueeze(inpainted, dim=0)
ret.append(inpainted)
ret = torch.cat(ret, dim=0)
return ret
def newblend(input, output, mask):
foreground = output
background = input
alpha = mask
foreground = cv2.multiply(alpha, foreground)
background = cv2.multiply(1.0 - alpha, background)
outImage = cv2.add(foreground, background)
return outImage
class Februusmodel(nn.Module):
def __init__(self, args):
super(Februusmodel, self).__init__()
self.base_model = generate_cls_model(model_name=args.model, num_classes=args.num_classes)
self.base_model.load_state_dict(result['model'])
self.base_model.eval()
self.inpaint_model = CompletionNetwork()
checkpoint = args.dataset + "_inpainting"
msg = self.inpaint_model.load_state_dict(torch.load(checkpoint, map_location='cuda'))
self.inpaint_model.eval()
print(msg)
self.data_mean, self.data_std = get_dataset_norm_stats(args.dataset)
self.data_mean = torch.tensor(self.data_mean).reshape(1,3,1,1).to(args.device)
self.data_std = torch.tensor(self.data_std).reshape(1,3,1,1).to(args.device)
self.data_sz = (args.input_height,args.input_width)
self.target_layers = [eval('self.base_model.{}'.format(args.cam_layer))]
self.gcam = eval(args.cam_method)(model=self.base_model, target_layers=self.target_layers, use_cuda=True)
self.MASK_COND = args.MASK_COND
self.device = args.device
self.mpv = self.data_mean
if args.dataset == 'gtsrb':
self.mpv = torch.tensor([0.33373367140503546, 0.3057189632961195, 0.316509230828686]).to(args.device)
self.mpv = self.mpv.view(1,3,1,1)
elif args.dataset == 'VGGFace2':
self.mpv = torch.tensor([0.5, 0.5, 0.5]).to(args.device)
self.mpv = self.mpv.view(1, 3, 1, 1)
def normalize(self, x, mean, std):
return (x-mean.to(x.device)) / std.to(x.device)
def unnormalize(self, x, mean, std):
return (x * std.to(x.device) + mean.to(x.device))
def data_normalize(self, x):
return self.normalize(x, self.data_mean, self.data_std)
def data_unnormalize(self, x):
return self.unnormalize(x, self.data_mean, self.data_std)
def forward(self, x):
outputs = self.base_model(x)
return outputs
def forward_inpaint(self, images):
cleanimgs = []
maskedimages = []
# GAN inpainted
# This is to apply Grad CAM to the load images
# --------------------------------------------
for j in range(len(images)):
image = images[j]
# image = self.februus_unnormalize(image) # unnormalize to [0 1] to feed into GAN
image = torch.unsqueeze(image, 0) # unsqueeze meaning adding 1D to the tensor
mask = self.gcam(image) # get the mask through GradCAM
cond_mask = mask >= self.MASK_COND
mask = cond_mask.astype(int)
# ---------------------------------------
mask = np.expand_dims(mask, axis=0) # add 1D to mask
# mask = np.expand_dims(mask, axis=0)
mask = torch.tensor(mask) # convert mask to tensor 1,1,32,32
mask = mask.type(torch.FloatTensor)
mask = mask.to(self.device)
x = self.data_unnormalize(image) # original test image
# inpaint
with torch.no_grad():
x_mask = x - x * mask + self.mpv * mask # generate the occluded input [0 1]
inputx = torch.cat((x_mask, mask), dim=1)
output = self.inpaint_model(inputx) # generate the output for the occluded input [0 1]
# image restoration
inpainted = poisson_blend_old(x_mask, output, mask) # this is GAN output [0 1]
inpainted = inpainted.to(self.device)
# store GAN output
cleanimgs.append(inpainted)
maskedimages.append(x_mask)
maskedimages = torch.cat(maskedimages)
maskedimages = self.data_normalize(maskedimages)
masked_outputs = self.base_model(maskedimages)
# this is tensor for GAN output
cleanimgs = torch.cat(cleanimgs)
cleanimgs = self.data_normalize(cleanimgs)
GAN_outputs = self.base_model(cleanimgs)
return (masked_outputs, GAN_outputs)
def get_args():
#set the basic parameter
parser = argparse.ArgumentParser()
parser.add_argument('--device', type=str, help='cuda, cpu')
parser.add_argument('--checkpoint_load', type=str)
parser.add_argument('--checkpoint_save', type=str)
parser.add_argument('--log', type=str)
parser.add_argument("--data_root", type=str)
parser.add_argument('--dataset', type=str, help='mnist, cifar10, gtsrb, celeba, tiny')
parser.add_argument("--num_classes", type=int)
parser.add_argument("--input_height", type=int)
parser.add_argument("--input_width", type=int)
parser.add_argument("--input_channel", type=int)
parser.add_argument('--epochs', type=int)
parser.add_argument('--batch_size', type=int)
parser.add_argument("--num_workers", type=float)
parser.add_argument('--lr', type=float)
parser.add_argument('--lr_scheduler', type=str, help='the scheduler of lr')
parser.add_argument('--poison_rate', type=float)
parser.add_argument('--target_type', type=str, help='all2one, all2all, cleanLabel')
parser.add_argument('--target_label', type=int)
parser.add_argument('--model', type=str, help='resnet18')
parser.add_argument('--seed', type=str, help='random seed')
parser.add_argument('--index', type=str, help='index of clean data')
parser.add_argument('--result_file', type=str, help='the location of result')
parser.add_argument('--yaml_path', type=str, default="./config/defense/ft/config.yaml", help='the path of yaml')
#set the parameter for the ft defense
parser.add_argument('--ratio', type=float, help='the ratio of clean data loader')
parser.add_argument('--inpaint_arch', type=str, default='inpaint_vit_base_patch16')
parser.add_argument('--inpaint_ckp', type=str, default='inpaint_visualize_vit_base.pth')
parser.add_argument('--inpaint_num', type=int, default=1)
parser.add_argument('--mask_ratio', type=float, default=0.75)
parser.add_argument('--MASK_COND', type=float, default=None)
parser.add_argument('--cam_layer', type=str, default='layer4[-1]')
parser.add_argument('--cam_method', type=str, default='GradCAM')
arg = parser.parse_args()
print(arg)
return arg
def defense(args, result,):
logFormatter = logging.Formatter(
fmt='%(asctime)s [%(levelname)-8s] [%(filename)s:%(lineno)d] %(message)s',
datefmt='%Y-%m-%d:%H:%M:%S',
)
logger = logging.getLogger()
if args.log_file_name is not None:
fileHandler = logging.FileHandler(os.getcwd() + args.log + '/' + args.log_file_name.split('/')[-1] + '.log')
else:
if args.log is not None and args.log != '':
fileHandler = logging.FileHandler(os.getcwd() + args.log + '/' + time.strftime("%Y_%m_%d_%H_%M_%S", time.localtime()) + '.log')
else:
fileHandler = logging.FileHandler(os.getcwd() + './log' + '/' + time.strftime("%Y_%m_%d_%H_%M_%S", time.localtime()) + '.log')
fileHandler.setFormatter(logFormatter)
logger.addHandler(fileHandler)
consoleHandler = logging.StreamHandler()
consoleHandler.setFormatter(logFormatter)
logger.addHandler(consoleHandler)
logger.setLevel(logging.INFO)
logging.info(pformat(args.__dict__))
fix_random(args.seed)
model = Februusmodel(args=args)
model.to(args.device)
tran = get_transform(args.dataset, *([args.input_height,args.input_width]) , train= False, norm=True)
x = result['bd_test']['x']
y = result['bd_test']['y']
data_bd_test = list(zip(x,y))
data_bd_testset = prepro_cls_DatasetBD(
full_dataset_without_transform=data_bd_test,
poison_idx=np.zeros(len(data_bd_test)), # one-hot to determine which image may take bd_transform
bd_image_pre_transform=None,
bd_label_pre_transform=None,
ori_image_transform_in_loading=tran,
ori_label_transform_in_loading=None,
add_details_in_preprocess=False,
)
data_bd_loader = torch.utils.data.DataLoader(data_bd_testset, batch_size=args.batch_size, num_workers=args.num_workers,drop_last=False, shuffle=False,pin_memory=True)
x = result['bd_test']['x']
y = result['bd_test']['original_targets']
data_bd_test_org = list(zip(x, y))
data_bd_testset_org = prepro_cls_DatasetBD(
full_dataset_without_transform=data_bd_test_org,
poison_idx=np.zeros(len(data_bd_test)),
bd_image_pre_transform=None,
bd_label_pre_transform=None,
ori_image_transform_in_loading=tran,
ori_label_transform_in_loading=None,
add_details_in_preprocess=False,
)
data_bd_loader_org = torch.utils.data.DataLoader(data_bd_testset_org, batch_size=args.batch_size,
num_workers=args.num_workers, drop_last=False, shuffle=False,
pin_memory=True)
tran = get_transform(args.dataset, *([args.input_height,args.input_width]) , train = False)
x = result['clean_test']['x']
y = result['clean_test']['y']
data_clean_test = list(zip(x,y))
data_clean_testset = prepro_cls_DatasetBD(
full_dataset_without_transform=data_clean_test,
poison_idx=np.zeros(len(data_clean_test)),
bd_image_pre_transform=None,
bd_label_pre_transform=None,
ori_image_transform_in_loading=tran,
ori_label_transform_in_loading=None,
add_details_in_preprocess=False,
)
data_clean_loader = torch.utils.data.DataLoader(data_clean_testset, batch_size=args.batch_size, num_workers=args.num_workers,drop_last=False, shuffle=False,pin_memory=True)
model.eval()
asr_acc = 0
for i, (inputs, labels) in enumerate(data_bd_loader):
inputs, labels = inputs.to(args.device), labels.to(args.device)
outputs = model(inputs)
pre_label = torch.max(outputs, dim=1)[1]
asr_acc += torch.sum(pre_label == labels)
original_ASR = asr_acc.item() / len(data_bd_loader.dataset) * 100
clean_acc = 0
for i, (inputs, labels) in enumerate(data_clean_loader):
inputs, labels = inputs.to(args.device), labels.to(args.device)
outputs = model(inputs)
pre_label = torch.max(outputs, dim=1)[1]
clean_acc += torch.sum(pre_label == labels)
original_ACC = clean_acc.item() / len(data_clean_loader.dataset) * 100
logging.info("original ACC is {} and original ASR is {}".format(original_ACC, original_ASR))
M = 2
clean_acc = [0] * M
cnt = [0] * M
for i, (inputs, labels) in enumerate(data_clean_loader):
inputs, labels = inputs.to(args.device), labels.to(args.device)
outputs = model.forward_inpaint(inputs)
for m in range(M):
pre_label = torch.max(outputs[m].detach().cpu(), dim=1)[1]
clean_acc[m] += torch.sum(pre_label == labels.cpu())
cnt[m] += len(labels)
for m in range(M):
inpaint_ACC = clean_acc[m].item() / len(data_clean_loader.dataset) * 100
logging.info("ACC w/ Februus is {} for method {}".format(inpaint_ACC, m))
asr_acc = [0] * M
cnt = [0] * M
all_preds = [[] for _ in range(M)]
for i, (inputs, labels) in enumerate(data_bd_loader_org):
inputs, labels = inputs.to(args.device), labels.to(args.device)
outputs = model.forward_inpaint(inputs)
for m in range(M):
pre_label = torch.max(outputs[m].detach().cpu(), dim=1)[1]
asr_acc[m] += torch.sum(pre_label == labels.cpu())
cnt[m] += len(labels)
all_preds[m].append(pre_label.detach())
for m in range(M):
inpaint_ACC_org = asr_acc[m].item() / len(data_bd_loader_org.dataset) * 100
logging.info("ACC on BAD IMAGES w/ Februus is {} for method {}".format(inpaint_ACC_org, m))
all_labels = []
for i, (inputs, labels) in enumerate(data_bd_loader):
all_labels.append(labels)
all_labels = torch.cat(all_labels)
for m in range(M):
inpaint_ASR = (torch.cat(all_preds[m])==all_labels).float().mean().item() * 100
logging.info("ASR w/ Februus is {} for method {}".format(inpaint_ASR, m))
if __name__ == '__main__':
### 1. basic setting: args
args = get_args()
with open(args.yaml_path, 'r') as stream:
config = yaml.safe_load(stream)
config.update({k:v for k,v in args.__dict__.items() if v is not None})
args.__dict__ = config
if args.dataset == "mnist":
args.num_classes = 10
args.input_height = 28
args.input_width = 28
args.input_channel = 1
elif args.dataset == "cifar10":
args.num_classes = 10
args.input_height = 32
args.input_width = 32
args.input_channel = 3
elif args.dataset == "cifar100":
args.num_classes = 100
args.input_height = 32
args.input_width = 32
args.input_channel = 3
elif args.dataset == "gtsrb":
args.num_classes = 43
args.input_height = 32
args.input_width = 32
args.input_channel = 3
elif args.dataset == "celeba":
args.num_classes = 8
args.input_height = 64
args.input_width = 64
args.input_channel = 3
elif args.dataset == "tiny":
args.num_classes = 200
args.input_height = 64
args.input_width = 64
args.input_channel = 3
elif "imagenet10" in args.dataset:
args.num_classes = 10
args.input_height = 224
args.input_width = 224
args.input_channel = 3
elif args.dataset == 'VGGFace2':
args.num_classes = 170
args.input_height = 224
args.input_width = 224
args.input_channel = 3
else:
raise Exception("Invalid Dataset")
save_path = '/record/' + args.result_file
if args.checkpoint_save is None:
args.checkpoint_save = save_path + '/record/defence/ft/'
if args.log is None:
args.log = save_path + '/saved/ft/'
else:
args.log_file_name = args.result_file + '_' + str(args.seed)
if not (os.path.exists(os.getcwd() + args.log)):
os.makedirs(os.getcwd() + args.log)
args.save_path = save_path
### 2. attack result(model, train data, test data)
result = load_attack_result(os.getcwd() + save_path + '/attack_result.pt', load_val=False, load_train=False)
### 3. ft defense:
defense(args, result)