Newer
Older
{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[]},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"}},"cells":[{"cell_type":"code","source":["import tensorflow as tf\n","import numpy as np\n","from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense, Dropout\n","from tensorflow.keras.models import Model\n","from tensorflow.keras.datasets import cifar10\n","from tensorflow.keras.utils import to_categorical\n","\n","from sklearn.model_selection import train_test_split\n","\n"],"metadata":{"id":"uG3R2ERwwYnS","executionInfo":{"status":"ok","timestamp":1702666988656,"user_tz":300,"elapsed":10391,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":1,"outputs":[]},{"cell_type":"code","source":["# Load Cifar10 dataset\n","(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n","\n","\n","# Concatenate train and test sets\n","x = np.concatenate((x_train, x_test))\n","y = np.concatenate((y_train, y_test))\n","\n","# Normalize the images\n","x = x.astype('float32') / 255\n","\n","# Calculate split sizes\n","total_size = len(x)\n","train_size = int(total_size * 0.70)\n","val_size = int(total_size * 0.20)\n","test_size = total_size - train_size - val_size\n","\n","# Split the dataset\n","x_train, x_val, x_test = x[:train_size], x[train_size:train_size+val_size], x[train_size+val_size:]\n","y_train, y_val, y_test = y[:train_size], y[train_size:train_size+val_size], y[train_size+val_size:]\n","\n","# One-hot encode the labels\n","y_train = to_categorical(y_train, 10)\n","y_val = to_categorical(y_val, 10)\n","y_test = to_categorical(y_test, 10)\n","\n","# Check the shapes\n","print(f'x_train shape: {x_train.shape}, y_train shape: {y_train.shape}')\n","print(f'x_val shape: {x_val.shape}, y_val shape: {y_val.shape}')\n","print(f'x_test shape: {x_test.shape}, y_test shape: {y_test.shape}')\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"f1HW9kHG5CG4","outputId":"e62b1f7b-ae8c-4adb-f69b-109075a9762f","executionInfo":{"status":"ok","timestamp":1702667011170,"user_tz":300,"elapsed":22533,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":2,"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n","170498071/170498071 [==============================] - 11s 0us/step\n","x_train shape: (42000, 32, 32, 3), y_train shape: (42000, 10)\n","x_val shape: (12000, 32, 32, 3), y_val shape: (12000, 10)\n","x_test shape: (6000, 32, 32, 3), y_test shape: (6000, 10)\n"]}]},{"cell_type":"code","source":["import matplotlib.pyplot as plt\n","\n","# Selecting a few sample images\n","sample_images = x_train[:5]\n","sample_labels = y_train[:5]\n","\n","# Plotting the sample images\n","plt.figure(figsize=(10, 2))\n","for i in range(len(sample_images)):\n"," plt.subplot(1, 5, i + 1)\n"," plt.imshow(sample_images[i], cmap='gray')\n"," #plt.title(f\"Label: {sample_labels[i]}\")\n"," plt.axis('off')\n","plt.show()\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":170},"id":"KlA-Ep0n55zr","outputId":"20cf768e-9aee-4d8b-9ebb-ce10f4390b67","executionInfo":{"status":"ok","timestamp":1702667011502,"user_tz":300,"elapsed":342,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":3,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 1000x200 with 5 Axes>"],"image/png":"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\n"},"metadata":{}}]},{"cell_type":"code","source":["from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, BatchNormalization\n","from tensorflow.keras.models import Sequential\n","\n","model = Sequential()\n","\n","model.add(Conv2D(32, (3, 3), activation='relu', padding='same', input_shape=(32, 32, 3)))\n","model.add(BatchNormalization())\n","model.add(Conv2D(32, (3, 3), activation='relu', padding='same'))\n","model.add(BatchNormalization())\n","model.add(MaxPooling2D(2, 2))\n","model.add(Dropout(0.2))\n","\n","model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))\n","model.add(BatchNormalization())\n","model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))\n","model.add(BatchNormalization())\n","model.add(MaxPooling2D(2, 2))\n","model.add(Dropout(0.3))\n","\n","model.add(Flatten())\n","model.add(Dense(512, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n","model.add(Dropout(0.5))\n","model.add(Dense(10, activation='softmax'))\n","\n","# Compile the model\n","adam = tf.keras.optimizers.Adam(learning_rate=0.001)\n","model.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])\n"],"metadata":{"id":"wXTVj63kMK-X","executionInfo":{"status":"ok","timestamp":1702667012430,"user_tz":300,"elapsed":930,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":4,"outputs":[]},{"cell_type":"code","source":["from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\n","\n","checkpoint = ModelCheckpoint(\"./model1.h5\", monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n","\n","early_stopping = EarlyStopping(monitor = 'val_loss',\n"," min_delta = 0,\n"," patience = 3,\n"," verbose = 1,\n"," restore_best_weights = True\n"," )\n","\n","reduce_learningrate = ReduceLROnPlateau(monitor = 'val_loss',\n"," factor = 0.2,\n"," patience = 3,\n"," verbose = 1,\n"," min_delta = 0.0001)\n","\n","callbacks_list = [early_stopping, checkpoint, reduce_learningrate]\n"],"metadata":{"id":"genQOGw-Mec8","executionInfo":{"status":"ok","timestamp":1702667012430,"user_tz":300,"elapsed":2,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":5,"outputs":[]},{"cell_type":"code","source":["def add_backdoor(x, y, target_label):\n"," backdoor_pattern = np.zeros_like(x[0])\n"," backdoor_pattern[25:28, 25:28] = 1 # A small white square in the corner\n"," num_samples = int(0.2 * x.shape[0]) # 20% of the dataset\n","\n"," for i in range(num_samples):\n"," x[i] += backdoor_pattern\n"," y[i] = to_categorical(target_label, 10)\n","\n"," return x, y\n","\n","# Insert backdoor\n","x_train_b, y_train_b = add_backdoor(x_train, y_train, target_label=0)\n"],"metadata":{"id":"zZfluLjP55sb","executionInfo":{"status":"ok","timestamp":1702667040244,"user_tz":300,"elapsed":317,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":8,"outputs":[]},{"cell_type":"code","source":["# Train the model on poisoned data\n","history = model.fit(x_train_b, y_train_b, batch_size=128, epochs=10, validation_data=(x_val, y_val))\n","\n","# Evaluate on clean data\n","loss, accuracy = model.evaluate(x_test, y_test)\n","print(f\"Clean test data accuracy: {accuracy}\")\n","\n","# Evaluate on backdoored data\n","x_test_backdoored, _ = add_backdoor(x_test, y_test, target_label=1)\n","loss, backdoor_accuracy = model.evaluate(x_test_backdoored, y_test)\n","print(f\"Backdoored test data accuracy: {backdoor_accuracy}\")\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"MSggOFxWCuNE","outputId":"f50f864f-dfc7-430d-b361-187a1725709d","executionInfo":{"status":"ok","timestamp":1702670096341,"user_tz":300,"elapsed":3054380,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":9,"outputs":[{"output_type":"stream","name":"stdout","text":["Epoch 1/10\n","329/329 [==============================] - 293s 890ms/step - loss: 1.0714 - accuracy: 0.8678 - val_loss: 2.7121 - val_accuracy: 0.2042\n","Epoch 2/10\n","329/329 [==============================] - 290s 882ms/step - loss: 0.6199 - accuracy: 0.8907 - val_loss: 1.6801 - val_accuracy: 0.4733\n","Epoch 3/10\n","329/329 [==============================] - 301s 915ms/step - loss: 0.4871 - accuracy: 0.9032 - val_loss: 1.5801 - val_accuracy: 0.5103\n","Epoch 4/10\n","329/329 [==============================] - 317s 963ms/step - loss: 0.4413 - accuracy: 0.9117 - val_loss: 1.4013 - val_accuracy: 0.5540\n","Epoch 5/10\n","329/329 [==============================] - 297s 904ms/step - loss: 0.4104 - accuracy: 0.9219 - val_loss: 1.4106 - val_accuracy: 0.5742\n","Epoch 6/10\n","329/329 [==============================] - 304s 926ms/step - loss: 0.4037 - accuracy: 0.9266 - val_loss: 1.5482 - val_accuracy: 0.5341\n","Epoch 7/10\n","329/329 [==============================] - 294s 895ms/step - loss: 0.3896 - accuracy: 0.9302 - val_loss: 1.2972 - val_accuracy: 0.6018\n","Epoch 8/10\n","329/329 [==============================] - 301s 915ms/step - loss: 0.3887 - accuracy: 0.9343 - val_loss: 1.1795 - val_accuracy: 0.6539\n","Epoch 9/10\n","329/329 [==============================] - 302s 918ms/step - loss: 0.3841 - accuracy: 0.9350 - val_loss: 1.1625 - val_accuracy: 0.6568\n","Epoch 10/10\n","329/329 [==============================] - 292s 888ms/step - loss: 0.3815 - accuracy: 0.9374 - val_loss: 1.3100 - val_accuracy: 0.6295\n","188/188 [==============================] - 11s 58ms/step - loss: 1.2928 - accuracy: 0.6232\n","Clean test data accuracy: 0.6231666803359985\n","188/188 [==============================] - 10s 53ms/step - loss: 9.8473 - accuracy: 0.4935\n","Backdoored test data accuracy: 0.4934999942779541\n"]}]},{"cell_type":"code","source":["x_test_backdoored = add_backdoor(x_test)"],"metadata":{"id":"knfu6zKROJgo","executionInfo":{"status":"aborted","timestamp":1702667031646,"user_tz":300,"elapsed":2,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Selecting a few sample images\n","backdoored_images = x_test_backdoored[:5]\n","\n","# Plotting the sample images\n","plt.figure(figsize=(10, 2))\n","for i in range(len(backdoored_images)):\n"," plt.subplot(1, 5, i + 1)\n"," plt.imshow(backdoored_images[i], cmap='gray')\n"," #plt.title(f\"Label: {backdoored_labels[i]}\")\n"," plt.axis('off')\n","plt.show()"],"metadata":{"id":"y56Pp7bzOVj1","executionInfo":{"status":"aborted","timestamp":1702667031646,"user_tz":300,"elapsed":2,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Plotting training and validation accuracy\n","plt.figure(figsize=(8, 4))\n","plt.plot(history.history['accuracy'], label='Training Accuracy')\n","plt.plot(history.history['val_accuracy'], label='Validation Accuracy')\n","plt.title('Training and Validation Accuracy')\n","plt.xlabel('Epoch')\n","plt.ylabel('Accuracy')\n","plt.legend()\n","plt.show()"],"metadata":{"id":"nVxaQGzw7EEi","executionInfo":{"status":"aborted","timestamp":1702667031646,"user_tz":300,"elapsed":2,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["from sklearn.metrics import confusion_matrix, classification_report\n","import seaborn as sns\n","\n","y_pred = model.predict(x_test)\n","y_pred_classes = np.argmax(y_pred, axis=1)\n","y_true = np.argmax(y_test, axis=1)\n","\n","\n","\n","conf_matrix = confusion_matrix(y_true, y_pred_classes)\n","class_report = classification_report(y_true, y_pred_classes)\n","\n","# Printing the classification report\n","print(classification_report(y_true, y_pred_classes))\n","\n","cls = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\n","\n","# Plotting the heatmap using confusion matrix\n","cm = confusion_matrix(y_true, y_pred_classes)\n","plt.figure(figsize = (8, 5))\n","sns.heatmap(cm, annot = True, fmt = '.0f', xticklabels = cls, yticklabels = cls)\n","plt.ylabel('Actual')\n","plt.xlabel('Predicted')\n","plt.show()"],"metadata":{"id":"6Yug_BgJdR21","executionInfo":{"status":"aborted","timestamp":1702667031646,"user_tz":300,"elapsed":2,"user":{"displayName":"TAMARA STUGAN","userId":"09145662160487804942"}}},"execution_count":null,"outputs":[]}]}