Evaluation/noise parameter analysis mnist.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1154ae38",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt \n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torchvision\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import gc\n",
    "from captum.attr import *\n",
    "import quantus\n",
    "from torch.utils.data import DataLoader\n",
    "import gc\n",
    "import torchvision.transforms as transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9015c61f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "db1d31a5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Device configuration\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b285e957",
   "metadata": {},
   "outputs": [],
   "source": [
    "#for natural and adversarial LeNet Model \n",
    "class LeNet_normal(torch.nn.Module):\n",
    "    \"\"\"Network architecture from: https://github.com/ChawDoe/LeNet5-MNIST-PyTorch.\"\"\"\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.conv_1 = torch.nn.Conv2d(1, 6, 5)\n",
    "        self.pool_1 = torch.nn.MaxPool2d(2, 2)\n",
    "        self.relu_1 = torch.nn.ReLU()\n",
    "        self.conv_2 = torch.nn.Conv2d(6, 16, 5)\n",
    "        self.pool_2 = torch.nn.MaxPool2d(2, 2)\n",
    "        self.relu_2 = torch.nn.ReLU()\n",
    "        self.fc_1 = torch.nn.Linear(256, 120)\n",
    "        self.relu_3 = torch.nn.ReLU()\n",
    "        self.fc_2 = torch.nn.Linear(120, 84)\n",
    "        self.relu_4 = torch.nn.ReLU()\n",
    "        self.fc_3 = torch.nn.Linear(84, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.pool_1(self.relu_1(self.conv_1(x)))\n",
    "        x = self.pool_2(self.relu_2(self.conv_2(x)))\n",
    "        x = x.view(x.shape[0], -1)\n",
    "        x = self.relu_3(self.fc_1(x))\n",
    "        x = self.relu_4(self.fc_2(x))\n",
    "        x = self.fc_3(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6d45233a",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_mnist_model(path):\n",
    "    model = LeNet_normal()\n",
    "    model.to(device)\n",
    "    model.load_state_dict(torch.load(path))\n",
    "    model.to('cuda')\n",
    "    model.train(False)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "fd1f181d",
   "metadata": {},
   "outputs": [],
   "source": [
    "modelpath = \"/data/virtual environments/adv detection by robustness/adv_detection/Adaptive attacks/Models/MNIST/mnist_model.pth\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "cc077361",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LeNet_normal(\n",
       "  (conv_1): Conv2d(1, 6, kernel_size=(5, 5), stride=(1, 1))\n",
       "  (pool_1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  (relu_1): ReLU()\n",
       "  (conv_2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))\n",
       "  (pool_2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  (relu_2): ReLU()\n",
       "  (fc_1): Linear(in_features=256, out_features=120, bias=True)\n",
       "  (relu_3): ReLU()\n",
       "  (fc_2): Linear(in_features=120, out_features=84, bias=True)\n",
       "  (relu_4): ReLU()\n",
       "  (fc_3): Linear(in_features=84, out_features=10, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "normal_model = load_mnist_model(modelpath)\n",
    "normal_model.to(device)\n",
    "normal_model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ec4cc932",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4d271c22",
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_noise(x_batch, y_batch, spread):\n",
    "    new_x_batch = []\n",
    "    for x in x_batch:\n",
    "        x = x.data.cpu().numpy()\n",
    "        stdev = spread * (np.max(x)-np.min(x))\n",
    "        noise = np.random.normal(0, stdev, x.shape).astype(np.float32)\n",
    "        x_plus_noise = x + noise\n",
    "        x_plus_noise = np.clip(x_plus_noise, 0, 1)\n",
    "        x_plus_noise = torch.from_numpy(x_plus_noise).cpu()\n",
    "        new_x_batch.append(x_plus_noise)\n",
    "    new_batch = torch.stack(new_x_batch).to(device)\n",
    "    return new_batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "d5d1b860",
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_metrics_benign(adv_path, normal_model, noise_parameter): \n",
    "    \n",
    "    print(\"Computing metrics for {} for benign\")\n",
    "    \n",
    "    npobj = np.load(adv_path)\n",
    "    adaptive_image = npobj['b_images']\n",
    "    adaptive_label = npobj['b_labels']\n",
    "    \n",
    "    \n",
    "    #attribution robustness\n",
    "    attribution_gaussian1 = []\n",
    "    #logit robustness\n",
    "    logit_gaussian1 = []\n",
    "   \n",
    "    \n",
    "    images, labels = torch.from_numpy(adaptive_image), torch.from_numpy(adaptive_label)\n",
    "    #images, labels = images.to(device), labels.to(device)\n",
    "    \n",
    "    end = len(adaptive_label)\n",
    "    if end > 1000:\n",
    "        end = 1000\n",
    "    \n",
    "    for i in range(0, end, 2):\n",
    "        \n",
    "        images_adv, y_pred_adv = images[i:i+2], labels[i:i+2]\n",
    "        images_adv, y_pred_adv = images_adv.to(device), y_pred_adv.to(device)\n",
    "        \n",
    "        x_logits = normal_model(images_adv)\n",
    "        gc.collect()\n",
    "        torch.cuda.empty_cache()\n",
    "        \n",
    "        #approach: attribution and logit robustness\n",
    "        a_batch = quantus.explain(\n",
    "            model=normal_model, inputs=images_adv, targets=y_pred_adv, **{\"method:\": \"IntegratedGradient\", \"device\": device})\n",
    "        \n",
    "        \n",
    "        gaussian_noisy_images_1 = make_noise(images_adv, y_pred_adv, spread = noise_parameter)\n",
    "        gaussian_logits_1 = normal_model(gaussian_noisy_images_1)\n",
    "        \n",
    "        \n",
    "        diff1 = torch.norm(x_logits-gaussian_logits_1,p=1, dim=1) \n",
    "        logit_gaussian1.extend(diff1.detach().cpu().numpy())\n",
    "        \n",
    "        \n",
    "        \n",
    "        a_batch_gaussian1 = quantus.explain(\n",
    "        model=normal_model, inputs=gaussian_noisy_images_1, targets=y_pred_adv, **{\"method:\": \"IntegratedGradient\", \"device\": device})\n",
    "        \n",
    "            \n",
    "        for a, b in zip(a_batch, a_batch_gaussian1):\n",
    "            c = np.linalg.norm(a.flatten()-b.flatten(),ord=1 )\n",
    "            attribution_gaussian1.append(c)\n",
    "            \n",
    "       \n",
    "        \n",
    "    df = pd.DataFrame([\n",
    "            attribution_gaussian1,\n",
    "            logit_gaussian1], index = [\n",
    "            \"Gaussian1 attribution\", \n",
    "            \"Gaussian1 logit robusntess\"\n",
    "                    ])\n",
    "            \n",
    "    path = \"adaptive_Benign.csv\"\n",
    "    df.to_csv(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "dcec9fef",
   "metadata": {},
   "outputs": [],
   "source": [
    "def return_thresholds(data_path, model):\n",
    "    noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07, 0.10]\n",
    "    final_auc_list = []\n",
    "    final_tpr_list = [] \n",
    "    final_fpr_list = [] \n",
    "    \n",
    "    for noise_p in noise_:\n",
    "        print('computing evaluations for noise {} '.format(noise_p))\n",
    "        compute_metrics_benign(data_path, model, noise_p)\n",
    "        df_cifar = pd.read_csv(\"adaptive_Benign.csv\")\n",
    "        attr_gaussian3 = df_cifar.iloc[0].values.flatten().tolist()[1:]\n",
    "        logit_gaussian3 = df_cifar.iloc[1].values.flatten().tolist()[1:]\n",
    "        print('Minimum attr:', min(attr_gaussian3))\n",
    "        print('Maximum attr:', max(attr_gaussian3))\n",
    "        print('Minimum logit:', min(logit_gaussian3))\n",
    "        print('Maximum logit:', max(logit_gaussian3))\n",
    "        print('---------')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "654a200c",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing evaluations for noise 0.001 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 4.2724423e-07\n",
      "Maximum attr: 233.75424\n",
      "Minimum logit: 0.009063244\n",
      "Maximum logit: 0.9191055\n",
      "---------\n",
      "computing evaluations for noise 0.002 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 9.154901e-07\n",
      "Maximum attr: 212.17314\n",
      "Minimum logit: 0.013617158\n",
      "Maximum logit: 2.181942\n",
      "---------\n",
      "computing evaluations for noise 0.003 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 1.3299286e-06\n",
      "Maximum attr: 283.53784\n",
      "Minimum logit: 0.03462434\n",
      "Maximum logit: 2.8588233\n",
      "---------\n",
      "computing evaluations for noise 0.004 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 9.536743e-07\n",
      "Maximum attr: 386.27377\n",
      "Minimum logit: 0.06614268\n",
      "Maximum logit: 4.640523\n",
      "---------\n",
      "computing evaluations for noise 0.005 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 2.1662563e-06\n",
      "Maximum attr: 295.55814\n",
      "Minimum logit: 0.051588535\n",
      "Maximum logit: 5.4401026\n",
      "---------\n",
      "computing evaluations for noise 0.006 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 7.373281e-06\n",
      "Maximum attr: 289.08148\n",
      "Minimum logit: 0.066069484\n",
      "Maximum logit: 5.6926484\n",
      "---------\n",
      "computing evaluations for noise 0.007 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 5.5441633e-06\n",
      "Maximum attr: 317.14545\n",
      "Minimum logit: 0.08151543\n",
      "Maximum logit: 7.149475\n",
      "---------\n",
      "computing evaluations for noise 0.008 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 1.1920094\n",
      "Maximum attr: 330.32172\n",
      "Minimum logit: 0.0786984\n",
      "Maximum logit: 8.593302\n",
      "---------\n",
      "computing evaluations for noise 0.009 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 1.5233655\n",
      "Maximum attr: 375.2081\n",
      "Minimum logit: 0.087350905\n",
      "Maximum logit: 7.735525\n",
      "---------\n",
      "computing evaluations for noise 0.01 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 3.261967\n",
      "Maximum attr: 417.31732\n",
      "Minimum logit: 0.16022587\n",
      "Maximum logit: 10.101432\n",
      "---------\n",
      "computing evaluations for noise 0.03 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 29.528852\n",
      "Maximum attr: 639.78326\n",
      "Minimum logit: 0.22613561\n",
      "Maximum logit: 28.745396\n",
      "---------\n",
      "computing evaluations for noise 0.05 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 49.73881\n",
      "Maximum attr: 1046.8511\n",
      "Minimum logit: 0.9036616\n",
      "Maximum logit: 53.25215\n",
      "---------\n",
      "computing evaluations for noise 0.07 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 87.49532\n",
      "Maximum attr: 1035.0134\n",
      "Minimum logit: 1.0340619\n",
      "Maximum logit: 61.495457\n",
      "---------\n",
      "computing evaluations for noise 0.1 \n",
      "Computing metrics for {} for benign\n",
      "Minimum attr: 86.32097\n",
      "Maximum attr: 1311.8965\n",
      "Minimum logit: 1.3409908\n",
      "Maximum logit: 92.08646\n",
      "---------\n"
     ]
    }
   ],
   "source": [
    "data_path = 'adv samples/MNIST/FGSM/0.03137254901960784eps.npz'\n",
    "return_thresholds(data_path, normal_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "538c2ac7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "k=[]\n",
    "l=[]\n",
    "m=[]\n",
    "n=[]\n",
    "\n",
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,1.5, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,200, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "0d7e6a87",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,1.7, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,200, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "57f09e1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,2, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,200, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "d41823cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,3, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,200, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "f04fa354",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,4, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,200, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "81d33a5b",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,6, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,260, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "fe19f5f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,7, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,260, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "9aae6c79",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.1,7, 10).tolist()\n",
    "\n",
    "m1=10*[4.2724423e-05]\n",
    "n1= np.linspace(5,260, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "a1aacb17",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.5,8, 10).tolist()\n",
    "\n",
    "m1=10*[0.1]\n",
    "n1= np.linspace(5,360, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "462bc73a",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(0.5,20, 10).tolist()\n",
    "\n",
    "m1=10*[0.1]\n",
    "n1= np.linspace(5,500, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "2f50f3a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(1.0,40, 10).tolist()\n",
    "\n",
    "m1=10*[0.1]\n",
    "n1= np.linspace(5,800, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "73e24738",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(1.0,50, 10).tolist()\n",
    "\n",
    "m1=10*[0.1]\n",
    "n1= np.linspace(5,800, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "3423de0a",
   "metadata": {},
   "outputs": [],
   "source": [
    "k1=10*[0.005]\n",
    "l1= np.linspace(1.0,50, 10).tolist()\n",
    "\n",
    "m1=10*[0.1]\n",
    "n1= np.linspace(5,800, 10).tolist()\n",
    "\n",
    "k.append(k1)\n",
    "l.append(l1)\n",
    "m.append(m1)\n",
    "n.append(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "c4580b9f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10, 10, 10, 10)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(k1), len(l1), len(m1), len(n1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "22b56324",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(k)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "77980925",
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_metrics_adv(adv_path, normal_model, noise_parameter): \n",
    "    \n",
    "    print(\"Computing metrics for {} for adv\")\n",
    "    npobj = np.load(adv_path)\n",
    "    adaptive_image = npobj['a_images']\n",
    "    adaptive_label = npobj['a_labels']\n",
    "    \n",
    "    #attribution robustness\n",
    "    attribution_gaussian1 = []\n",
    "    \n",
    "    #logit robustness\n",
    "    logit_gaussian1 = []\n",
    "  \n",
    "    images, labels = torch.from_numpy(adaptive_image), torch.from_numpy(adaptive_label)\n",
    "    #images, labels = images.to(device), labels.to(device)\n",
    "    \n",
    "    end = len(adaptive_label)\n",
    "    if end > 1000:\n",
    "        end = 1000\n",
    "    \n",
    "    for i in range(0, end, 2):\n",
    "        \n",
    "        images_adv, y_pred_adv = images[i:i+2], labels[i:i+2]\n",
    "        images_adv, y_pred_adv = images_adv.to(device), y_pred_adv.to(device)\n",
    "        \n",
    "        x_logits = normal_model(images_adv)\n",
    "        gc.collect()\n",
    "        torch.cuda.empty_cache()\n",
    "        \n",
    "        #approach: attribution and logit robustness\n",
    "        a_batch = quantus.explain(\n",
    "            model=normal_model, inputs=images_adv, targets=y_pred_adv, **{\"method:\": \"IntegratedGradient\", \"device\": device})\n",
    "        \n",
    "        gaussian_noisy_images_1 = make_noise(images_adv, y_pred_adv, spread = noise_parameter)\n",
    "        gaussian_logits_1 = normal_model(gaussian_noisy_images_1)\n",
    "        \n",
    "        diff1 = torch.norm(x_logits-gaussian_logits_1,p=1, dim=1) \n",
    "        logit_gaussian1.extend(diff1.detach().cpu().numpy())\n",
    "     \n",
    "        \n",
    "        a_batch_gaussian1 = quantus.explain(\n",
    "        model=normal_model, inputs=gaussian_noisy_images_1, targets=y_pred_adv, **{\"method:\": \"IntegratedGradient\", \"device\": device})\n",
    "        \n",
    "        for a, b in zip(a_batch, a_batch_gaussian1):\n",
    "            c = np.linalg.norm(a.flatten()-b.flatten(),ord=1 )\n",
    "            attribution_gaussian1.append(c)\n",
    "            \n",
    "        \n",
    "    df = pd.DataFrame([\n",
    "            \n",
    "            attribution_gaussian1,\n",
    "            logit_gaussian1], index = [\n",
    "            \"Gaussian1 attribution\",\n",
    "            \"Gaussian1 logit robusntess\"\n",
    "                    ])\n",
    "            \n",
    "    path = \"adaptive_Adv.csv\"\n",
    "    df.to_csv(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "0e4c3e37",
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_TPR(adv1, a, b, adv2, c, d): \n",
    "    TN=0\n",
    "    FN=0\n",
    "    FP = 0 \n",
    "    TP=0\n",
    "    \n",
    "    for value1, value2 in zip(adv1, adv2): \n",
    "        if value1<a or value1>b:\n",
    "            TP += 1\n",
    "        else:\n",
    "            if value2<c or value2>d:\n",
    "                TP+=1\n",
    "            else: \n",
    "                FN+=1\n",
    "    \n",
    "    return (TP/(TP+FN))*100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "ed11de69",
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_FPR(ap2a, k, l, ap2b, m, n): \n",
    "    TN=0\n",
    "    FN=0\n",
    "    FP=0 \n",
    "    TP=0\n",
    "    \n",
    "    for value6, value7 in zip(ap2a,ap2b):\n",
    "        if value6<k or value6>l:\n",
    "            FP +=1\n",
    "        else:\n",
    "            if value7<m or value7>n:\n",
    "                FP +=1\n",
    "\n",
    "    return (FP/(len(ap2a)))*100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "dfc203be",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sklearn\n",
    "from sklearn.metrics import roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "8f6b6290",
   "metadata": {},
   "outputs": [],
   "source": [
    "def return_auc(adv_path, model, k, l, m, n, noise_p):\n",
    "    #logitgaussian3\n",
    "    k=k\n",
    "    l=l\n",
    "\n",
    "    #attrgaussian3\n",
    "    m=m\n",
    "    n=n\n",
    "\n",
    "    print('computing evaluations for noise {} '.format(noise_p))\n",
    "    compute_metrics_benign(adv_path, model, noise_p)\n",
    "    compute_metrics_adv(adv_path, model, noise_p)\n",
    "    df_cifar = pd.read_csv(\"adaptive_Benign.csv\")\n",
    "    attr_gaussian3 = df_cifar.iloc[0].values.flatten().tolist()[1:]\n",
    "    logit_gaussian3 = df_cifar.iloc[1].values.flatten().tolist()[1:]\n",
    "    fpr_results =[]\n",
    "    for t1,t2,t3,t4 in zip(k,l,m,n):\n",
    "        FPR = compute_FPR(logit_gaussian3, t1,t2, attr_gaussian3,t3,t4)\n",
    "        fpr_results.append(FPR/100)\n",
    "    df_pgd_eps1 = pd.read_csv(\"adaptive_Adv.csv\")\n",
    "    attr_gaussian3_eps1 = df_pgd_eps1.iloc[0].values.flatten().tolist()[1:]\n",
    "    logit_gaussian3_eps1 = df_pgd_eps1.iloc[1].values.flatten().tolist()[1:]\n",
    "    tpr_results =[]\n",
    "    for t1,t2,t3,t4 in zip(k,l,m,n):\n",
    "        TPR = compute_TPR(logit_gaussian3_eps1, t1,t2, attr_gaussian3_eps1,t3,t4)\n",
    "        tpr_results.append(TPR/100)\n",
    "    auc_ = sklearn.metrics.auc(fpr_results, tpr_results)\n",
    "    \n",
    "    return auc_, fpr_results, tpr_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "21c3b36f",
   "metadata": {},
   "outputs": [],
   "source": [
    "adv_path = 'adv samples/MNIST/FGSM/0.03137254901960784eps.npz'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "64dc55bd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(13, 14)"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(k), len(noise_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "3093faac",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing evaluations for noise 0.001 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.001 is 0.5759265.\n",
      "---\n",
      "computing evaluations for noise 0.002 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.002 is 0.834647.\n",
      "---\n",
      "computing evaluations for noise 0.003 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.003 is 0.9288885.\n",
      "---\n",
      "computing evaluations for noise 0.004 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.004 is 0.9608345000000001.\n",
      "---\n",
      "computing evaluations for noise 0.005 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.005 is 0.9743725000000001.\n",
      "---\n",
      "computing evaluations for noise 0.006 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.006 is 0.9750610000000002.\n",
      "---\n",
      "computing evaluations for noise 0.007 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.007 is 0.9765129999999999.\n",
      "---\n",
      "computing evaluations for noise 0.008 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.008 is 0.9646035000000001.\n",
      "---\n",
      "computing evaluations for noise 0.009 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.009 is 0.9614715000000001.\n",
      "---\n",
      "computing evaluations for noise 0.01 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.01 is 0.9295735000000002.\n",
      "---\n",
      "computing evaluations for noise 0.03 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.03 is 0.6435515.\n",
      "---\n",
      "computing evaluations for noise 0.05 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.05 is 0.4697295.\n",
      "---\n",
      "computing evaluations for noise 0.07 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.07 is 0.39177599999999996.\n",
      "---\n"
     ]
    },
    {
     "ename": "IndexError",
     "evalue": "list index out of range",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[37], line 7\u001b[0m\n\u001b[1;32m      4\u001b[0m fgsm_fpr_list \u001b[38;5;241m=\u001b[39m [] \n\u001b[1;32m      6\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m i \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mrange\u001b[39m(\u001b[38;5;28mlen\u001b[39m(noise_)): \n\u001b[0;32m----> 7\u001b[0m     auc, fpr, tpr \u001b[38;5;241m=\u001b[39m return_auc(adv_path, normal_model, \u001b[43mk\u001b[49m\u001b[43m[\u001b[49m\u001b[43mi\u001b[49m\u001b[43m]\u001b[49m, l[i], m[i], n[i], noise_[i])\n\u001b[1;32m      8\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mauc for noise \u001b[39m\u001b[38;5;132;01m{}\u001b[39;00m\u001b[38;5;124m is \u001b[39m\u001b[38;5;132;01m{}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;241m.\u001b[39mformat(noise_[i], auc))\n\u001b[1;32m      9\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m---\u001b[39m\u001b[38;5;124m'\u001b[39m)\n",
      "\u001b[0;31mIndexError\u001b[0m: list index out of range"
     ]
    }
   ],
   "source": [
    "\n",
    "noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07, 0.10]\n",
    "fgsm_auc_list = []\n",
    "fgsm_tpr_list = [] \n",
    "fgsm_fpr_list = [] \n",
    "\n",
    "for i in range(len(noise_)): \n",
    "    auc, fpr, tpr = return_auc(adv_path, normal_model, k[i], l[i], m[i], n[i], noise_[i])\n",
    "    print('auc for noise {} is {}.'.format(noise_[i], auc))\n",
    "    print('---')\n",
    "    fgsm_auc_list.append(auc)\n",
    "    fgsm_tpr_list.append(tpr)\n",
    "    fgsm_fpr_list.append(fpr)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf04ab3c",
   "metadata": {},
   "source": [
    "# PGD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "5f6ccd87",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing evaluations for noise 0.001 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.001 is 0.5690000000000001.\n",
      "---\n",
      "computing evaluations for noise 0.002 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.002 is 0.8500000000000001.\n",
      "---\n",
      "computing evaluations for noise 0.003 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.003 is 0.933.\n",
      "---\n",
      "computing evaluations for noise 0.004 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.004 is 0.9630000000000001.\n",
      "---\n",
      "computing evaluations for noise 0.005 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.005 is 0.9720000000000002.\n",
      "---\n",
      "computing evaluations for noise 0.006 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.006 is 0.9849875000000001.\n",
      "---\n",
      "computing evaluations for noise 0.007 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.007 is 0.9909565.\n",
      "---\n",
      "computing evaluations for noise 0.008 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.008 is 0.988905.\n",
      "---\n",
      "computing evaluations for noise 0.009 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.009 is 0.995861.\n",
      "---\n",
      "computing evaluations for noise 0.01 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.01 is 0.9980144999999999.\n",
      "---\n",
      "computing evaluations for noise 0.03 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.03 is 0.937696.\n",
      "---\n",
      "computing evaluations for noise 0.05 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.05 is 0.7763595000000001.\n",
      "---\n",
      "computing evaluations for noise 0.07 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.07 is 0.6801505.\n",
      "---\n"
     ]
    }
   ],
   "source": [
    "adv_path1 = 'adv samples/MNIST/PGD/0.03137254901960784eps.npz'\n",
    "noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07]\n",
    "pgd_auc_list = []\n",
    "pgd_tpr_list = [] \n",
    "pgd_fpr_list = [] \n",
    "\n",
    "for i in range(len(noise_)): \n",
    "    auc, fpr, tpr = return_auc(adv_path1, normal_model, k[i], l[i], m[i], n[i], noise_[i])\n",
    "    print('auc for noise {} is {}.'.format(noise_[i], auc))\n",
    "    print('---')\n",
    "    pgd_auc_list.append(auc)\n",
    "    pgd_tpr_list.append(tpr)\n",
    "    pgd_fpr_list.append(fpr)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5759050a",
   "metadata": {},
   "source": [
    "# BIM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "faea2b8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing evaluations for noise 0.001 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.001 is 0.3043245000000001.\n",
      "---\n",
      "computing evaluations for noise 0.002 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.002 is 0.4726715000000001.\n",
      "---\n",
      "computing evaluations for noise 0.003 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.003 is 0.5239130000000001.\n",
      "---\n",
      "computing evaluations for noise 0.004 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.004 is 0.535415.\n",
      "---\n",
      "computing evaluations for noise 0.005 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.005 is 0.5592189999999999.\n",
      "---\n",
      "computing evaluations for noise 0.006 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.006 is 0.5819444999999999.\n",
      "---\n",
      "computing evaluations for noise 0.007 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.007 is 0.5521815.\n",
      "---\n",
      "computing evaluations for noise 0.008 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.008 is 0.578907.\n",
      "---\n",
      "computing evaluations for noise 0.009 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.009 is 0.5728765.\n",
      "---\n",
      "computing evaluations for noise 0.01 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.01 is 0.545545.\n",
      "---\n",
      "computing evaluations for noise 0.03 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.03 is 0.43291750000000007.\n",
      "---\n",
      "computing evaluations for noise 0.05 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.05 is 0.31690699999999994.\n",
      "---\n",
      "computing evaluations for noise 0.07 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.07 is 0.2838645.\n",
      "---\n"
     ]
    }
   ],
   "source": [
    "adv_path1 = 'adv samples/MNIST/BIM/0.03137254901960784eps.npz'\n",
    "noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07]\n",
    "bim_auc_list = []\n",
    "bim_tpr_list = [] \n",
    "bim_fpr_list = [] \n",
    "\n",
    "for i in range(len(noise_)): \n",
    "    auc, fpr, tpr = return_auc(adv_path1, normal_model, k[i], l[i], m[i], n[i], noise_[i])\n",
    "    print('auc for noise {} is {}.'.format(noise_[i], auc))\n",
    "    print('---')\n",
    "    bim_auc_list.append(auc)\n",
    "    bim_tpr_list.append(tpr)\n",
    "    bim_fpr_list.append(fpr)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e0bedf4",
   "metadata": {},
   "source": [
    "# CW"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "2fd8e0ed",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing evaluations for noise 0.001 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.001 is 0.28535399999999994.\n",
      "---\n",
      "computing evaluations for noise 0.002 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.002 is 0.33338600000000007.\n",
      "---\n",
      "computing evaluations for noise 0.003 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.003 is 0.364725.\n",
      "---\n",
      "computing evaluations for noise 0.004 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.004 is 0.3654275000000001.\n",
      "---\n",
      "computing evaluations for noise 0.005 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.005 is 0.404619.\n",
      "---\n",
      "computing evaluations for noise 0.006 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.006 is 0.4021495.\n",
      "---\n",
      "computing evaluations for noise 0.007 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.007 is 0.406508.\n",
      "---\n",
      "computing evaluations for noise 0.008 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.008 is 0.40232299999999993.\n",
      "---\n",
      "computing evaluations for noise 0.009 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.009 is 0.417614.\n",
      "---\n",
      "computing evaluations for noise 0.01 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.01 is 0.44904249999999996.\n",
      "---\n",
      "computing evaluations for noise 0.03 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.03 is 0.3338060000000001.\n",
      "---\n",
      "computing evaluations for noise 0.05 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.05 is 0.29408199999999995.\n",
      "---\n",
      "computing evaluations for noise 0.07 \n",
      "Computing metrics for {} for benign\n",
      "Computing metrics for {} for adv\n",
      "auc for noise 0.07 is 0.28193100000000004.\n",
      "---\n"
     ]
    }
   ],
   "source": [
    "adv_path1 = 'adv samples/MNIST/CW/00.15eps.npz'\n",
    "noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07]\n",
    "cw_auc_list = []\n",
    "cw_tpr_list = [] \n",
    "cw_fpr_list = [] \n",
    "\n",
    "for i in range(len(noise_)): \n",
    "    auc, fpr, tpr = return_auc(adv_path1, normal_model, k[i], l[i], m[i], n[i], noise_[i])\n",
    "    print('auc for noise {} is {}.'.format(noise_[i], auc))\n",
    "    print('---')\n",
    "    cw_auc_list.append(auc)\n",
    "    cw_tpr_list.append(tpr)\n",
    "    cw_fpr_list.append(fpr)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "fcc53d0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "auc=[]\n",
    "for a,b,c,d in zip(fgsm_auc_list, pgd_auc_list, bim_auc_list, cw_auc_list):\n",
    "    z = (a+b+c+d)/4 \n",
    "    auc.append(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "bf7963fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "noise_ = [0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01,0.03, 0.05, 0.07]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "6145b11a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fe9b858db80>]"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.xlabel('Noise spread parameter')\n",
    "plt.ylabel('AUC')\n",
    "plt.plot(noise_, auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "16274c90",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "plt.plot(noise_, fgsm_auc_list, label=\"FGSM\")\n",
    "plt.plot(noise_, pgd_auc_list, label=\"PGD\")\n",
    "plt.plot(noise_, bim_auc_list, label=\"BIM\")\n",
    "plt.plot(noise_, cw_auc_list, label=\"CW\")\n",
    "plt.xlabel('Noise spread parameter')\n",
    "plt.ylabel('AUC')\n",
    "\n",
    "plt.legend()\n",
    "\n",
    "plt.savefig('parameteranalysismnist.pdf', format='pdf', bbox_inches='tight')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7141168a",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "adv_detection",
   "language": "python",
   "name": "adv_detection"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}