Steganografiya quyidagi sohalarda qo'llaniladi, lekin ular bilan cheklanmaydi

else:
# Case where inputs are wrongly classified
classes = self.args.nbr_classes
list_labels = [k for k in range(0, classes) if k != x_true]
returned_results.append(random.choice(list_labels))

return returned_results
else:
return self.model.predict(X_test)

def fit(self, X_train, y_train):
""" Train the model on watermarked data

Args:
X_train (array): Input data
y_train (array): Label data (0 / 1)

Returns:
ownership (dict): Watermark triggers information

"""
# Check type of the model
if type(self.model) in self.Classifiers:
# Generate triggers
ownership, X_train, y_train = self.generate_trigger(
X_train,
y_train,
mode='CLASSIFICATION')
elif type(self.model) in self.Regressors:
# Generate triggers
ownership, X_train, y_train = self.generate_trigger(
X_train,
y_train,
mode='REGRESSION')
else:
raise NotImplementedError()
# Train the model

selected_q = self.train_step(ownership, X_train, y_train)
ownership['selected_q'] = selected_q
# Store encrypted triggers
if self.encryption:
return self.encrypt(ownership)
else:
return ownership

def encrypt(self, ownership):
""" Store the watermark in encrypted fashion.

Args:
ownership (dict): Watermark triggers information

Returns:
encrypted_trigger (dict): Encrypted triggers information

"""
# Initiating the keys and the triggers(crypted)
keys = []