/
task_utils.py
174 lines (136 loc) · 5.72 KB
/
task_utils.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
import json
import os
import gen_utils
import keras
from tensorflow.keras.utils import to_categorical
from keras.layers import Conv2D, MaxPooling2D, Dense, Flatten, Activation, Dropout, BatchNormalization
from keras.models import Model
from keras.models import Sequential
from keras.regularizers import l2
class Task(object):
def __init__(self, name, load_clean=False, test_only=False):
self.name = name
self.load_clean = load_clean
if name == "cifar":
self.num_classes = 10
self.img_shape = (32, 32, 3)
self.feature_layer_name = 'flatten'
else:
raise Exception("Not implement")
if test_only:
self.X_test, self.Y_test = load_dataset(name, test_only=test_only)
else:
self.X_train, self.Y_train, self.X_test, self.Y_test = load_dataset(name, test_only=test_only)
self.number_train = len(self.X_train)
if load_clean is not None:
self.model = self.get_model(load_clean=self.load_clean)
self.bottleneck_model = self.build_bottleneck_model()
def build_bottleneck_model(self):
self.bottleneck_model = build_bottleneck_model(self.model, self.feature_layer_name)
return self.bottleneck_model
def get_model(self, load_clean=False):
model = get_model(self.name, load_clean=load_clean)
return model
def get_student_model(self):
assert self.load_clean or self.name == 'cifar'
student_model = get_student_model(self.bottleneck_model, self.num_classes)
return student_model
def get_model(dataset, load_clean=False):
assert dataset == 'cifar'
if load_clean:
# model = keras.models.load_model("models/cifar_cifar2_model.h5")
model = keras.models.load_model(
"/home/shansixioing/forensic/wide_resnets_keras-master/models/cifar_clean.h5")
else:
# model = create_model()
model = get_cifar_model()
return model
def get_cifar_model(softmax=True):
layers = [
Conv2D(32, (3, 3), padding='same', input_shape=(32, 32, 3)), # 0
Activation('relu'), # 1
BatchNormalization(), # 2
Conv2D(32, (3, 3), padding='same'), # 3
Activation('relu'), # 4
BatchNormalization(), # 5
MaxPooling2D(pool_size=(2, 2)), # 6
Conv2D(64, (3, 3), padding='same'), # 7
Activation('relu'), # 8
BatchNormalization(), # 9
Conv2D(64, (3, 3), padding='same'), # 10
Activation('relu'), # 11
BatchNormalization(), # 12
MaxPooling2D(pool_size=(2, 2)), # 13
Conv2D(128, (3, 3), padding='same'), # 14
Activation('relu'), # 15
BatchNormalization(), # 16
Conv2D(128, (3, 3), padding='same'), # 17
Activation('relu'), # 18
BatchNormalization(), # 19
MaxPooling2D(pool_size=(2, 2)), # 20
Flatten(), # 21
Dropout(0.2), # 22
Dense(2048, kernel_regularizer=l2(0.01), bias_regularizer=l2(0.01)), # 23
Activation('relu'), # 24
BatchNormalization(), # 25
Dense(512, kernel_regularizer=l2(0.01), bias_regularizer=l2(0.01)), # 27
Activation('relu'), # 28
BatchNormalization(), # 29
Dense(10), # 31
]
model = Sequential()
for layer in layers:
model.add(layer)
if softmax:
model.add(Activation('softmax'))
return model
def load_dataset(dataset, test_only=False):
if dataset == "cifar":
from keras.datasets import cifar10
(X_train, Y_train), (X_test, Y_test) = cifar10.load_data()
if not test_only:
X_train = X_train / 255.0
Y_train = to_categorical(Y_train, 10)
X_test = X_test / 255.0
Y_test = to_categorical(Y_test, 10)
if test_only:
return X_test, Y_test
return X_train, Y_train, X_test, Y_test
def load_attack(config_name, load_clean=False):
config = json.load(open(os.path.join("configs/", config_name + ".json"), "r"))
dataset = config['dataset']
task = Task(dataset, load_clean=load_clean, test_only=True)
res_file = gen_utils.pickle_read(f"results/{dataset}_{config_name}_res.p")
target_ls = res_file['target_ls']
injected_X = res_file['injected_X']
injected_Y = res_file['injected_Y']
X_test = res_file['X_test']
Y_test = res_file['Y_test']
is_backdoor_ls = res_file['injected']
if "inject_ratio" not in res_file and "number_poison" in res_file: # clean label case
injected_X_test = res_file['target_image']
injected_Y_test = res_file['target_test_Y']
else:
injected_X_test = res_file['injected_X_test']
injected_Y_test = res_file['injected_Y_test']
target_label = target_ls[0]
task.X_test = X_test
task.Y_test = Y_test
number_train = len(injected_X)
return dataset, task, injected_X, injected_Y, X_test, Y_test, injected_X_test, injected_Y_test, is_backdoor_ls, number_train, task.feature_layer_name, task.num_classes, target_label
def get_student_model(bottleneck_model, num_classes):
for l in bottleneck_model.layers:
l.trainable = False
x = bottleneck_model.layers[-1].output
x = Dense(num_classes, name='logit')(x)
x = Activation('softmax', name='act')(x)
model = Model(bottleneck_model.input, x)
opt = keras.optimizers.Adam()
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
return model
def build_bottleneck_model(model, layer_name):
bottleneck_model = keras.models.Model(model.input, model.get_layer(layer_name).output)
bottleneck_model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
return bottleneck_model