Merge pull request #47 from Critical-Infrastructure-Systems-Lab/dev

Preparing version 0.5.0

Author: afmurillo

dhalsim/network_attacks/adversarial_models/ctown_attacker_scaler.gz

@@ -0,0 +1 @@
+���ʫ���h������ۮ�
������ ��������
(��労����
2

dhalsim/network_attacks/adversarial_models/ctown_generator_100_percent/fingerprint.pb

@@ -0,0 +1,8 @@
+
+�2root"_tf_keras_network*�2{"name": "generator", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": false, "class_name": "Functional", "config": {"name": "generator", "trainable": true, "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 39]}, "dtype": "float32", "sparse": false, "ragged": false, "name": "input_1"}, "name": "input_1", "inbound_nodes": []}, {"class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense", "inbound_nodes": [[["input_1", 0, 0, {}]]]}, {"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_1", "inbound_nodes": [[["dense", 0, 0, {}]]]}, {"class_name": "Dense", "config": {"name": "dense_2", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_2", "inbound_nodes": [[["dense_1", 0, 0, {}]]]}, {"class_name": "Dense", "config": {"name": "dense_3", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_3", "inbound_nodes": [[["dense_2", 0, 0, {}]]]}], "input_layers": [["input_1", 0, 0]], "output_layers": [["dense_3", 0, 0]]}, "shared_object_id": 13, "input_spec": [{"class_name": "InputSpec", "config": {"dtype": null, "shape": {"class_name": "__tuple__", "items": [null, 39]}, "ndim": 2, "max_ndim": null, "min_ndim": null, "axes": {}}}], "build_input_shape": {"class_name": "TensorShape", "items": [null, 39]}, "is_graph_network": true, "full_save_spec": {"class_name": "__tuple__", "items": [[{"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 39]}, "float32", "input_1"]}], {}]}, "save_spec": {"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 39]}, "float32", "input_1"]}, "keras_version": "2.13.1", "backend": "tensorflow", "model_config": {"class_name": "Functional", "config": {"name": "generator", "trainable": true, "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 39]}, "dtype": "float32", "sparse": false, "ragged": false, "name": "input_1"}, "name": "input_1", "inbound_nodes": [], "shared_object_id": 0}, {"class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 1}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 2}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense", "inbound_nodes": [[["input_1", 0, 0, {}]]], "shared_object_id": 3}, {"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 4}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 5}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_1", "inbound_nodes": [[["dense", 0, 0, {}]]], "shared_object_id": 6}, {"class_name": "Dense", "config": {"name": "dense_2", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 7}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 8}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_2", "inbound_nodes": [[["dense_1", 0, 0, {}]]], "shared_object_id": 9}, {"class_name": "Dense", "config": {"name": "dense_3", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 10}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 11}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_3", "inbound_nodes": [[["dense_2", 0, 0, {}]]], "shared_object_id": 12}], "input_layers": [["input_1", 0, 0]], "output_layers": [["dense_3", 0, 0]]}}, "training_config": {"loss": "mean_squared_error", "metrics": null, "weighted_metrics": null, "loss_weights": null, "optimizer_config": {"class_name": "Custom>Adam", "config": {"name": "Adam", "weight_decay": null, "clipnorm": null, "global_clipnorm": null, "clipvalue": null, "use_ema": false, "ema_momentum": 0.99, "ema_overwrite_frequency": null, "jit_compile": false, "is_legacy_optimizer": false, "learning_rate": 6.25000029685907e-05, "beta_1": 0.9, "beta_2": 0.999, "epsilon": 1e-07, "amsgrad": false}}}}2

+�
root.layer-0"_tf_keras_input_layer*�{"class_name": "InputLayer", "name": "input_1", "dtype": "float32", "sparse": false, "ragged": false, "batch_input_shape": {"class_name": "__tuple__", "items": [null, 39]}, "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 39]}, "dtype": "float32", "sparse": false, "ragged": false, "name": "input_1"}}2

+�root.layer_with_weights-0"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 1}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 2}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["input_1", 0, 0, {}]]], "shared_object_id": 3, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 39}}, "shared_object_id": 15}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 39]}}2

+�root.layer_with_weights-1"_tf_keras_layer*�{"name": "dense_1", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 4}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 5}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["dense", 0, 0, {}]]], "shared_object_id": 6, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 19}}, "shared_object_id": 16}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 19]}}2

+�root.layer_with_weights-2"_tf_keras_layer*�{"name": "dense_2", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense_2", "trainable": true, "dtype": "float32", "units": 19, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 7}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 8}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["dense_1", 0, 0, {}]]], "shared_object_id": 9, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 39}}, "shared_object_id": 17}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 39]}}2

+�root.layer_with_weights-3"_tf_keras_layer*�{"name": "dense_3", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense_3", "trainable": true, "dtype": "float32", "units": 39, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotNormal", "config": {"seed": 12345}, "shared_object_id": 10}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 11}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["dense_2", 0, 0, {}]]], "shared_object_id": 12, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 19}}, "shared_object_id": 18}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 19]}}2

+�
\root.keras_api.metrics.0"_tf_keras_metric*�
{"class_name": "Mean", "name": "loss", "dtype": "float32", "config": {"name": "loss", "dtype": "float32"}, "shared_object_id": 19}2

dhalsim/network_attacks/adversarial_models/ctown_generator_100_percent/keras_metadata.pb

@@ -12,9 +12,7 @@
 
 from dhalsim.network_attacks.utilities import launch_arp_poison, restore_arp
 from dhalsim.network_attacks.synced_attack import SyncedAttack
-from evasion_attacks.Adversarial_Attacks.Black_Box_Attack import adversarial_AE
-
-from tensorflow.keras.models import Model, load_model
+#from tensorflow.keras.models import Model, load_model
 
 
 class Error(Exception):

dhalsim/network_attacks/adversarial_models/ctown_generator_100_percent/saved_model.pb

@@ -3,6 +3,7 @@
 from tensorflow.keras.optimizers import Adam
 from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau
 from sklearn.preprocessing import MinMaxScaler
+import joblib
 from tensorflow.keras.initializers import glorot_normal
 
 import pandas as pd
@@ -13,6 +14,8 @@
 from sklearn.preprocessing import MinMaxScaler
 from sklearn.model_selection import train_test_split
 
+import joblib
+
 # This module is based on the implementation by Alessandro Erba, original is found here:
 # https://github.com/scy-phy/ICS-Evasion-Attacks/blob/master/Adversarial_Attacks/Black_Box_Attack/adversarial_AE.py
 
@@ -37,7 +40,10 @@ def preprocess_physical(self, path):
 
         return a_pd
 
-    def train_model(self):
+    def train_model(self, training_path):
+
+        self.init_generator(training_path)
+
         # fit of the scaler is done at __init__
         ben_data = self.physical_pd
         ben_data[self.sensor_cols] = self.attacker_scaler.transform(ben_data[self.sensor_cols])
@@ -54,8 +60,48 @@ def train_model(self):
                             verbose=2,
                             validation_data=(x_ben_test, x_ben_test))
 
+        # print('Printing data used for training')
+        ben_data.to_csv('trained_data.csv', index=False)
+
+    # Loads the scaler originally used to train the model
+    def load_scaler(self, scaler_path):        
+        #scaler_path = Path.cwd()/'attacker_scaler.gz'
+        
+        print('Loading scaler at', str(scaler_path))
+        self.attacker_scaler = joblib.load(str(scaler_path))
+        print('Scaler loaded')
+
+    # Saves the model and the scaler used to train the model
     def save_model(self, filename):
-        self.generator.save(filename)
+        print('saving trained model at: ', str(filename))
+        self.generator.save(str(model_path))
+
+        scaler_path = Path.cwd()
+        print('saved scaler model at: ', filename)
+        joblib.dump(self.attacker_scaler, 'ctown_attacker_scaler.gz')
+        
+    def init_generator(self, training_path):
+        # Load and preprocess training data
+        training_path = Path(__file__).parent/training_path/'training_data.csv'
+        # print('Reading training data from: ' + str(training_path))
+        self.physical_pd = self.preprocess_physical(training_path)
+
+        # Adversarial model for concealment
+        # toDo: Ask about this parameter
+        hide_layers = 39
+        self.hide_layers = hide_layers
+        self.generator_layers = [self.feature_dims,
+                                int(self.hide_layers / 2),
+                                self.hide_layers,
+                                int(self.hide_layers / 2), self.feature_dims]
+
+        optimizer = Adam(lr=0.001)
+        # Build the generator
+        self.generator = self.build_generator()
+        self.generator.compile(optimizer=optimizer, loss='mean_squared_error')
+
+        self.attacker_scaler = MinMaxScaler()
+        self.transform_fit_scaler()
 
     def build_generator(self):
         input = Input(shape=(self.feature_dims,))
@@ -85,40 +131,18 @@ def fix_sample(self, gen_examples):
 
     def predict(self, received_values_df):
         print('Attempting to predict concealment values')
+        # print('Features received to predict: ' + str(received_values_df.columns))
+        # print('Features received to train: ' + str(self.sensor_cols))
+
         gen_examples = self.generator.predict(self.attacker_scaler.transform(received_values_df))
         gen_examples = self.fix_sample(pd.DataFrame(columns=self.sensor_cols,
                                                     data=self.attacker_scaler.inverse_transform(gen_examples)))
 
         return gen_examples
 
-    def __init__(self, a_path):
-        # Load and preprocess training data
-        training_path = Path(__file__).parent/a_path/'training_data.csv'
-        print('Reading training data from: ' + str(training_path))
-        self.physical_pd = self.preprocess_physical(training_path)
-
-        # Adversarial model for concealment
-        # toDo: Ask about this parameter
-        hide_layers = 39
+    def __init__(self, features_list):
 
-        self.sensor_cols = [col for col in self.physical_pd.columns if
+        # Initialize model with features list
+        self.sensor_cols = [col for col in features_list if
                             col not in ['Unnamed: 0', 'iteration', 'timestamp', 'Attack']]
-
         self.feature_dims = len(self.sensor_cols)
-
-        self.hide_layers = hide_layers
-        self.generator_layers = [self.feature_dims,
-                                 int(self.hide_layers / 2),
-                                 self.hide_layers,
-                                 int(self.hide_layers / 2), self.feature_dims]
-
-        optimizer = Adam(lr=0.001)
-        # Build the generator
-        self.generator = self.build_generator()
-        self.generator.compile(optimizer=optimizer, loss='mean_squared_error')
-
-        self.attacker_scaler = MinMaxScaler()
-        self.transform_fit_scaler()
-
-
-