def main():
# tf.logging.set_verbosity(tf.logging.INFO)
# instantiate blackbox and substitute model
# instantiate blackbox and substitute model
forward_model = load_model()
# backward_model1 = create_fmodel_18()
backward_model2 = create_fmodel_ALP()
backward_model3 = create_fmodel_ALP1000()
# print(backward_model1[0])
# instantiate differntiable composite model
# (predictions from blackbox, gradients from substitute)
model = CompositeModel(
forward_model=forward_model,
backward_models=[backward_model2, backward_model3],
weights = [0.5, 0.5])
predictor = tp.OfflinePredictor(tp.PredictConfig(
model=SaliencyModel(),
session_init=tp.get_model_loader("resnet_v1_50.ckpt"),
input_names=['image'],
output_names=['saliency']))
for (file_name, image, label) in read_images():
pos_salience = find_salience(predictor, image)
adversarial = run_attack(model, image, label, pos_salience)
store_adversarial(file_name, adversarial)
attack_complete()
def main():
loader = TinyImageNetLoader()
forward_model = load_model()
backward_model1 = create_fmodel_ALP()
backward_model2 = create_fmodel_ALP1000()
model = CompositeModel(forward_model=forward_model,
backward_models=[backward_model1, backward_model2],
weights=[0.3, 0.7])
i = 0
total_sum = 0.0
prev_avg = 0.0
for (file_name, image, label) in read_images():
is_boundary = True
is_adam = True
adversarial = run_attack(loader, forward_model, image, label)
# Calculate both Adam and Boundary for first 5 images.
if i < 5:
adversarial_adam = run_attack2(model, image, label, None)
if adversarial is not None:
error1 = distance(adversarial, image)
else:
error1 = 100000.0
is_boundary = False
if adversarial_adam is not None:
error_adam = distance(adversarial_adam, image)
else:
error_adam = 200000.0
is_adam = False
if is_adam and error1 - error_adam > 0.0:
adversarial = adversarial_adam
if is_adam or is_boundary:
i += 1
total_sum += min(error1, error_adam)
else:
if adversarial is not None:
error1 = distance(adversarial, image)
prev_avg = total_sum / i
i += 1
total_sum += error1
else:
error1 = 100000.0
if error1 > 25.0 or error1 > prev_avg or adversarial is None:
adversarial_adam = run_attack2(model, image, label, None)
if adversarial_adam is not None:
error_adam = distance(adversarial_adam, image)
else:
error_adam = 200000.0
if error1 - error_adam > 0.0:
adversarial = adversarial_adam
store_adversarial(file_name, adversarial)
# Announce that the attack is complete
# NOTE: In the absence of this call, your submission will timeout
# while being graded.
attack_complete()
def main():
# tf.logging.set_verbosity(tf.logging.INFO)
# instantiate blackbox and substitute model
# instantiate blackbox and substitute model
forward_model = load_model()
backward_model1 = create_fmodel_18()
backward_model2 = create_fmodel_ALP()
# print(backward_model1[0])
# instantiate differntiable composite model
# (predictions from blackbox, gradients from substitute)
model = CompositeModel(forward_model=forward_model,
backward_models=[backward_model1, backward_model2])
for (file_name, image, label) in read_images():
adversarial = run_attack(model, image, label)
store_adversarial(file_name, adversarial)
attack_complete()
def main():
loader = TinyImageNetLoader()
forward_model = load_model()
backward_model1 = create_fmodel_ALP()
backward_model2 = create_fmodel_ALP1000()
model = CompositeModel(forward_model=forward_model,
backward_models=[backward_model1, backward_model2],
weights=[0.5, 0.5])
for (file_name, image, label) in read_images():
adversarial = run_attack(loader, forward_model, image, label)
if adversarial is None:
adversarial = run_attack2(model, image, label, None)
store_adversarial(file_name, adversarial)
# Announce that the attack is complete
# NOTE: In the absence of this call, your submission will timeout
# while being graded.
attack_complete()
def main():
loader = TinyImageNetLoader()
forward_model = load_model()
backward_model1 = create_fmodel_ALP()
backward_model2 = create_fmodel_ALP1000()
model = CompositeModel(forward_model=forward_model,
backward_models=[backward_model1, backward_model2],
weights=[0.5, 0.5])
predictor = tp.OfflinePredictor(
tp.PredictConfig(model=SaliencyModel(),
session_init=tp.get_model_loader("resnet_v1_50.ckpt"),
input_names=['image'],
output_names=['saliency']))
for (file_name, image, label) in read_images():
adversarial = run_attack(loader, forward_model, image, label)
if adversarial is None:
pos_salience = find_salience(predictor, image)
adversarial = run_attack2(model, image, label, pos_salience)
store_adversarial(file_name, adversarial)
# Announce that the attack is complete
# NOTE: In the absence of this call, your submission will timeout
# while being graded.
attack_complete()
