trinhudo / phy_security_noma_fd_dnn

Enhancing PHY-Security of FD-Enabled NOMA Systems Using Jamming and User Selection: Performance Analysis and DNN-based Evaluation

Paper

K. Shim, T. N. Do, T. -V. Nguyen, D. B. d. Costa and B. An, "Enhancing PHY-Security of FD-Enabled NOMA Systems Using Jamming and User Selection: Performance Analysis and DNN-based Evaluation," in IEEE Internet of Things Journal, vol. 8, no. 24, pp. 17476-17494, Dec., 2021.

DNN-based SOP Prediction Demo

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from tensorflow import keras
from sklearn.metrics import mean_squared_error
from tensorflow.keras.callbacks import ReduceLROnPlateau
#from keras.callbacks import ReduceLROnPlateau
from sklearn.model_selection import train_test_split
from keras.layers import BatchNormalization
from tensorflow.keras.layers import Dense, Dropout, \
    Activation, Flatten, Conv1D, MaxPooling1D

NUM_EPOCHS = 30 # The number of epochs for training.
BATCH_SIZE = 200 #  

data = pd.read_csv('DataSet_FarUserMTS_1e5.csv')
data.shape
data.head()

x_in = data.drop(columns=['SOP']) 
# feature set
# get all data in each column except the column with title "SOP"

y_in = data['SOP'] 
# target variable (output)
# get data in the column with title "SOP"

x_train, x_test, y_train, y_test = train_test_split(x_in,y_in,
                                                    test_size = 0.1,
                                                    random_state = 0)

print(x_test)
print(y_test)

## build DNN model for regression using Keras Sequential API
# Input Layer
model = keras.models.Sequential() # a basic feed-forward model
# Multiple Hidden Layer
model.add(keras.layers.Dense(128, activation='relu', 
                             input_shape = (12,), 
                             kernel_initializer='normal'))
model.add(keras.layers.Dense(128, activation='relu', 
                             kernel_initializer='normal'))
model.add(keras.layers.Dense(128, activation='relu', 
                             kernel_initializer='normal'))
model.add(keras.layers.Dense(128, activation='relu', 
                             kernel_initializer='normal'))
model.add(keras.layers.Dense(128, activation='relu', 
                             kernel_initializer='normal'))
# Output Layer
model.add(keras.layers.Dense(1, activation='linear',
                             kernel_initializer='normal'))  
# our output layer 


# Setting for the DNN model training. 
reduce_lr = ReduceLROnPlateau(monitor = 'val_loss', 
                              patience = 1, 
                              factor = 0.8, 
                              min_lr = 1e-10) 
# for adaptive learning rate

optimizer = keras.optimizers.Adam(lr = 1e-2) 
# optimizer for update DNN model


model.compile(optimizer = optimizer,
              loss='mse',
              metrics=['mse'])

history = model.fit(x_train, y_train, 
                    validation_split=0.1,
                    epochs = NUM_EPOCHS,  
                    batch_size = BATCH_SIZE,
                    callbacks = [reduce_lr]
                    )

       Unnamed: 0  Unnamed: 1  Unnamed: 2  ...  Unnamed: 9  Unnamed: 10  Unnamed: 11
3582            4          35           5  ...         0.1          0.2          0.1
60498           4          55          10  ...         0.2          0.2          0.2
53227           4         -15           5  ...         0.2          0.1          0.1
21333           4          40          10  ...         0.1          0.2          0.1
3885            4          10          10  ...         0.1          0.2          0.2
...           ...         ...         ...  ...         ...          ...          ...
89555           4           5           5  ...         0.1          0.1          0.2
88135           4         -15           5  ...         0.1          0.1          0.2
51888           4          60          10  ...         0.1          0.2          0.2
51380           4           0          10  ...         0.1          0.1          0.1
67294           4         -15           5  ...         0.1          0.1          0.2

[10000 rows x 12 columns]
3582     0.86520
60498    0.99975
53227    0.96625
21333    0.80355
3885     0.08580
          ...   
89555    0.22685
88135    0.99705
51888    0.99650
51380    0.84415
67294    0.98175
Name: SOP, Length: 10000, dtype: float64
Epoch 1/30
405/405 [==============================] - 3s 6ms/step - loss: 0.0770 - mse: 0.0770 - val_loss: 0.0020 - val_mse: 0.0020
Epoch 2/30
405/405 [==============================] - 2s 6ms/step - loss: 0.0015 - mse: 0.0015 - val_loss: 4.9305e-04 - val_mse: 4.9305e-04
...
Epoch 29/30
405/405 [==============================] - 2s 6ms/step - loss: 1.0618e-05 - mse: 1.0618e-05 - val_loss: 1.0652e-05 - val_mse: 1.0652e-05
Epoch 30/30
405/405 [==============================] - 2s 6ms/step - loss: 1.0404e-05 - mse: 1.0404e-05 - val_loss: 1.0470e-05 - val_mse: 1.0470e-05

# plt.plot(history.history['mean_squared_error'])
# plt.plot(history.history['val_mean_squared_error'])

# If the above history plot does not work, pls use the following
plt.plot(history.history['mse'])
plt.plot(history.history['val_mse'])

plt.yscale('log')
plt.legend(loc='upper right')
plt.grid(True)
plt.ylabel('Estimator MSE')
plt.xlabel('Number of Epochs')
plt.legend(['Train', 'Validation'], loc='upper right')
plt.show()

# the saved trainined DNN model
print(model.summary())
model.save('Trained_DNN_FarUserMTS.h5')

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_6 (Dense)              (None, 128)               1664      
_________________________________________________________________
dense_7 (Dense)              (None, 128)               16512     
_________________________________________________________________
dense_8 (Dense)              (None, 128)               16512     
_________________________________________________________________
dense_9 (Dense)              (None, 128)               16512     
_________________________________________________________________
dense_10 (Dense)             (None, 128)               16512     
_________________________________________________________________
dense_11 (Dense)             (None, 1)                 129       
=================================================================
Total params: 67,841
Trainable params: 67,841
Non-trainable params: 0
_________________________________________________________________
None

# verify the trained model
new_model = keras.models.load_model('Trained_DNN_FarUserMTS.h5')
y_pred = new_model.predict(x_test)

print('===============================================')
print('Root Mean Squared Error:', 
      np.sqrt(mean_squared_error(y_test, y_pred))) 
print('===============================================')
print('END')

===============================================
Root Mean Squared Error: 0.00331421030026788
===============================================
END