Tutorial - RamBoAttack - CIFAR10 .ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "78193fe6",
   "metadata": {},
   "source": [
    "### 0. Import libraries & functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "5ef22921",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"]=\"PCI_BUS_ID\"\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"0\";\n",
    "#from pudb import set_trace\n",
    "import numpy as np\n",
    "import torch\n",
    "from torchvision import models\n",
    "\n",
    "import torch.backends.cudnn as cudnn\n",
    "import random\n",
    "from tqdm import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from utils_rb import *\n",
    "from ramboattack import RamBoAtt\n",
    "from HSJA_rb import HSJA\n",
    "from SignOPT_rb import OPT_attack_sign_SGD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3a988c7c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def imshow(img):\n",
    "    \n",
    "    npimg = img[0].cpu().numpy()\n",
    "    npimg = np.transpose(npimg,(1, 2, 0))\n",
    "    npimg = np.clip(npimg, 0, 1)\n",
    "    plt.imshow(npimg)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "06f31f62",
   "metadata": {},
   "outputs": [],
   "source": [
    "def perturbation_heat_map(xo,xa):\n",
    "    \n",
    "    fig_dims = (5, 5)\n",
    "    fig, ax = plt.subplots(figsize=fig_dims)\n",
    "    x=torch.abs(xo-xa).sum(dim=1).cpu()[0]\n",
    "    sns.heatmap(x,ax=ax,xticklabels=False, yticklabels=False,cbar = False)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e1b25425",
   "metadata": {},
   "source": [
    "### 1. Load data, pretrained model, evaluation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f57bef7f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "# a. Load dataset\n",
    "\n",
    "batch_size = 1\n",
    "dataset = 'cifar10' \n",
    "datapath = '../datasets/cifar10'\n",
    "testloader, testset = load_data(dataset,data_path=datapath,batch_size=batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "3dcdced9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# b. Load pre-trained model\n",
    "\n",
    "# 'resnet50' if pre-trained model from Pytorch. 'cifar10' if using pre-trained cifar10 model\n",
    "arch = 'cifar10' \n",
    "\n",
    "# None means using pre-traineded model from Pytorch or default path. Otherwise, please change model_path = '...'\n",
    "model_path = None \n",
    "\n",
    "# True means pre-trained model does \"not\" normalized data while training, \n",
    "# so no need to unnorm during intergerence (used for CIFAR10 model)\n",
    "\n",
    "if dataset == 'cifar10':\n",
    "    num_classes = 10\n",
    "    unnorm = True # True means pre-trained model does \"not\" normalized data while training.\n",
    "    \n",
    "net = load_model(arch,model_path)\n",
    "model_rb = PretrainedModel(net,dataset,unnorm)\n",
    "\n",
    "bounds = [0,1]\n",
    "model_ex = PytorchModel_ex(net, bounds, num_classes,dataset,unnorm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "0b63ed2e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# c. Load evaluation set\n",
    "targeted = True # True means targeted attack. False means untargeted attack.\n",
    "# 'balance', 'easyset'->imagenet or cifar10; \n",
    "# 'hardset'-> imagenet; \n",
    "# 'hardset_A','hardset_B','hardset_D' -> cifar10\n",
    "eval_set =  'hardset_B'\n",
    "\n",
    "ID_set = get_evalset(dataset,targeted,eval_set)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a59b1a64",
   "metadata": {},
   "source": [
    "### 2. HopSkipJump Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "a08eb5c1",
   "metadata": {},
   "outputs": [],
   "source": [
    "i = 2 #0,1,2,10,20, 50,123 # the sample i-th in the evaluation set\n",
    "query_limit = 50000\n",
    "D = np.zeros(query_limit+2000)\n",
    "nquery = 0\n",
    "o = ID_set[i,1] #oID\n",
    "\n",
    "# 0. select original image\n",
    "oimg, olabel = testset[o]\n",
    "oimg = torch.unsqueeze(oimg, 0).cuda()\n",
    "\n",
    "# 1. select starting image\n",
    "if targeted:\n",
    "    t = ID_set[i,3] #tID, 3 is index acrross dataset - 4 is sample index in a class (not accross dataset)\n",
    "    tlabel = ID_set[i,2]\n",
    "    timg, _ = testset[t]\n",
    "    timg = torch.unsqueeze(timg, 0).cuda()\n",
    "    y_targ = np.array([tlabel])\n",
    "else:\n",
    "    tlabel = None\n",
    "    y_targ = np.array([olabel])\n",
    "    \n",
    "# ============= 2. setup ==============\n",
    "if dataset == 'cifar10':\n",
    "    delta = 1e-2\n",
    "    len_T = 500\n",
    "\n",
    "constraint='l2'\n",
    "num_iterations=150\n",
    "gamma=1.0\n",
    "stepsize_search='geometric_progression'\n",
    "max_num_evals = 1e4\n",
    "init_num_evals=100\n",
    "verbose=True\n",
    "auto_terminate=False\n",
    "\n",
    "#seed = 0\n",
    "#np.random.seed(seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "9eea7ca0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 1, l2 distance 1.1526E+01 - nqry:139\n",
      "iteration: 11, l2 distance 2.8740E+00 - nqry:2808\n",
      "iteration: 21, l2 distance 1.5103E+00 - nqry:7076\n",
      "iteration: 31, l2 distance 1.5022E+00 - nqry:12494\n",
      "iteration: 41, l2 distance 1.5005E+00 - nqry:18824\n",
      "iteration: 51, l2 distance 1.2988E+00 - nqry:25906\n",
      "iteration: 61, l2 distance 1.2571E+00 - nqry:33665\n",
      "iteration: 71, l2 distance 1.2545E+00 - nqry:42100\n",
      "\n",
      "Break due to over query limit!\n",
      "\n"
     ]
    }
   ],
   "source": [
    "model_ex = PytorchModel_ex(net, bounds, num_classes,dataset,unnorm)\n",
    "module = HSJA(model_ex,constraint,num_iterations,gamma,stepsize_search,max_num_evals,init_num_evals, verbose,delta,len_T)\n",
    "if targeted:\n",
    "    adv, nqry, Dt = module.hsja(oimg.cpu().numpy(), y_targ, timg.cpu().numpy(),targeted,query_limit,auto_terminate)\n",
    "else:\n",
    "    timg = None\n",
    "    adv, nqry, Dt = module.hsja(oimg.cpu().numpy(), y_targ, timg,targeted,auto_terminate)\n",
    "    \n",
    "adv = torch.unsqueeze(torch.from_numpy(adv).float(), 0).cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0ad1a832",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Source image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adversarial Example:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Perturbation Heat Map:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Source image:')\n",
    "imshow(oimg)\n",
    "\n",
    "print('Starting image:')\n",
    "timg, _ = testset[t]\n",
    "timg = torch.unsqueeze(timg, 0).cuda()\n",
    "imshow(timg)\n",
    "\n",
    "print('Adversarial Example:')\n",
    "imshow(adv)\n",
    "\n",
    "print('Perturbation Heat Map:')\n",
    "perturbation_heat_map(oimg,adv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "97d354ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "DHsja = Dt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1af32ecc",
   "metadata": {},
   "source": [
    "### 3. Sign-OPT Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "f60ad951",
   "metadata": {},
   "outputs": [],
   "source": [
    "i = 2#0,1,2,10,20,50,123 # the sample i-th in the evaluation set\n",
    "query_limit = 50000\n",
    "D = np.zeros(query_limit+2000)\n",
    "nquery = 0\n",
    "o = ID_set[i,1] #oID\n",
    "\n",
    "# 0. select original image\n",
    "oimg, olabel = testset[o]\n",
    "oimg = torch.unsqueeze(oimg, 0).cuda()\n",
    "\n",
    "# 1. select starting image\n",
    "if targeted:\n",
    "    t = ID_set[i,3] #tID, 3 is index acrross dataset - 4 is sample index in a class (not accross dataset)\n",
    "    tlabel = ID_set[i,2]\n",
    "    timg, _ = testset[t]\n",
    "    timg = torch.unsqueeze(timg, 0).cuda()\n",
    "    y_targ = np.array([tlabel])\n",
    "else:\n",
    "    tlabel = None\n",
    "    y_targ = np.array([olabel])\n",
    "    \n",
    "# ============= 2. setup ==============\n",
    "alpha = 0.2\n",
    "beta = 0.001\n",
    "iterations = 5000\n",
    "distortion = None\n",
    "stopping = 0.0001\n",
    "auto_terminate=False\n",
    "k=200\n",
    "\n",
    "seed = 0\n",
    "#np.random.seed(seed)\n",
    "\n",
    "if dataset == 'cifar10':\n",
    "    delta = 1e-2\n",
    "    len_T = 2\n",
    "elif dataset == 'imagenet':\n",
    "    delta = 1\n",
    "    len_T = 4\n",
    "\n",
    "module = OPT_attack_sign_SGD(model_rb,k,delta,len_T,testset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "1e3ad997",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Targeted attack - Source: 1 and Target: 5 - pred y0: tensor([1], device='cuda:0') - pred yt: tensor([5], device='cuda:0')\n",
      "--------> Found distortion 11.6200\n",
      "==========> Found best distortion 11.6200 in 0.0350 seconds using 22 queries\n",
      "Iteration   1 distortion 9.2015 num_queries 373\n",
      "Iteration  11 distortion 3.9571 num_queries 4240\n",
      "Iteration  21 distortion 3.4252 num_queries 7691\n",
      "Iteration  31 distortion 3.1196 num_queries 11343\n",
      "Iteration  41 distortion 3.0244 num_queries 14772\n",
      "Iteration  51 distortion 2.9568 num_queries 18012\n",
      "Iteration  61 distortion 2.9155 num_queries 21405\n",
      "Iteration  71 distortion 2.8279 num_queries 25476\n",
      "Iteration  81 distortion 2.4322 num_queries 34157\n",
      "Iteration  91 distortion 2.3741 num_queries 40034\n",
      "Iteration 101 distortion 2.0139 num_queries 47068\n",
      "break due to query cnt > query limit 51310\n",
      "\n",
      "Adversarial Example Found Successfully: distortion 1.9837 target 5 queries 51310 LS queries 30088 \n",
      "Time: 100.1367 seconds\n"
     ]
    }
   ],
   "source": [
    "if targeted:\n",
    "    adv, nqry, Dt = module.attack_targeted(oimg, olabel, timg, tlabel, alpha, beta, iterations, query_limit, distortion, seed, stopping, auto_terminate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "9dfc452d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Source image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adversarial Example:\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD5CAYAAADhukOtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAevElEQVR4nO2da4xd13Xf/+uc++TMkMPh8CU+REti7QhxTCmMYjuOa0uRqxhGZNetIBsQlFS1giIGaiD9ILhA7QL94BS1DX8oXNC1EDlwZTuxXCuNkFiS48oWDNkjW6IelCySJkNRQ3L4Gs7Mnfte/XAvW0rd/z0jzswdxvv/Awje2Wv2Oevue9Y9d/b/rrXM3SGE+NUnW20HhBCDQcEuRCIo2IVIBAW7EImgYBciERTsQiRCYSmTzew2AF8CkAP47+7+udjvj4+P+9W7doWNEQXQvRMcn6/X6ZxWvUFttflZfi60qG2+HrYVsxKdM7p+jNqGRoaoLc+L1BaFrKO9+Sm9eZGJ3uUz2byo0BuRgS3mSPTJMSM/l0efdMTHiBtxIzlVzNgNDx/9h6M4ffp08GyXHexmlgP4rwBuBfAqgJ+a2cPu/iKbc/WuXXhqYiJsbPNz1RsXguMHXn6Zzjn20ivUtv+FH1Jby05R27MHjgfHt5SvpnM+fMfHqe1dN++ltpGhbdSWRS6DNlnHSsavtnrkQiwZP1ez3uTzCuF57ciHyW43/KYOAOViTm2W82N6O/zkus7f1NtF/kZrbRJlAEqx94gs8ibhYf+b/CnD58J+vPt330XnLOVj/E0ADrr7YXdvAvgGgNuXcDwhxAqylGDfBuDYJT+/2h8TQlyBrPgGnZnda2YTZjZxempqpU8nhCAsJdiPA9hxyc/b+2Ovw933ufted987vnHjEk4nhFgKSwn2nwLYbWZvMbMSgDsBPLw8bgkhlpvL3o1397aZfRLA36Envd3v7i8sOI+Mv/bLw3TOc/vDG/xHXttP55ycepraJi+8Rm3V9Ruo7V//0SeC47954610zo6rdlJbg22dAygQuREAGhm3VQtEBsz5bnDW4TvTnnNZsZKVqa2Vh3fqc/DjdQpcLo1Lb3w9WFZnlvFLv9jg2+CNCt+Nb7b5vFJkN77RCs8rRiSqToX4H1FdlqSzu/sjAB5ZyjGEEINB36ATIhEU7EIkgoJdiERQsAuRCAp2IRJhSbvxb5bZmWk88YPw5v2TP+Yy2sxcWCqbn+VzWg0ux/z223+f2j7wgX9Fbduv2xIcjyU01WpcTioWIwkcEakpi0g89WZYrsnKEemtzf1oc6UMnQJf42Yh7H8lkkhSiMhyrYyvY97gEmC7HPajDe675ZHXrM7XvhBZj/kWf97o1ILD7XaFz6kyWS7yOvOjCSF+lVCwC5EICnYhEkHBLkQiKNiFSISB7sZfuDCD7z/2g6DtlVcO0nnnz54Njt92y1vpnPfe8s+pbff176G2vMjrwjUaZAe0w3dh15T4Es+TckQAUGzzHeGsHCntRCoqNTxSTqnDd/4jG9OYL/IagMV2eIfcI2sVWw/McR9LZV4eqxJRLigZ92O2wJOGCh2uCuSRXfIOOZ8RZQUAikQmMVpzT3d2IZJBwS5EIijYhUgEBbsQiaBgFyIRFOxCJMJApbfaXA0TPw53hNlyDe+q8vF7/jA4/v5f+206Z2jTOLXlRS5pNPw8tWWdtcFxq3K5oxtpkZR7pH2V8ZemPMflny7pnBJLnskLfD1mi/y5lX0NtWUIa3Y1pg0CKHQjNfkitd+6hSq1TZPEm0qHy3VW5j4WPdKWy/jr0mnw+6qVwteI5fw1827YFmsZpTu7EImgYBciERTsQiSCgl2IRFCwC5EICnYhEmFJ0puZHQEwA6ADoO3ue2O/P7phAz561x8GbR/9yO/ReUNrNgfH25HspGbOpZo8IoPk5XXcRmQca0dqsbUiWUjFSBsnruJglkg1AFAhvsSyzeqRInqFnBstklBWmwuvVYlIcgAQKYUHi7So6kYkqszCB21kfE63ySXRYed14SJLjE4eyVQkcmS3w6/hcjV8sliXrOXQ2d/v7qeX4ThCiBVEH+OFSISlBrsD+J6ZPW1m9y6HQ0KIlWGpH+Pf4+7HzWwTgEfN7CV3f+LSX+i/CdwLAGMbeDtkIcTKsqQ7u7sf7/9/CsB3ANwU+J197r7X3fcOj4ws5XRCiCVw2cFuZkNmNnLxMYAPAHh+uRwTQiwvS/kYvxnAd6y3118A8D/c/W9jEzZuGMM9d98ZtOXGW//M+FzYMMfdzwpcWukYz1zK6jzzar4QzpQqNLgsFFFcUOhE2jhVuI9rcq7xMFHOKnxOmStNUfmn4/yYQxmbx7PNug1+DbSL/LUuxDLYisPhOVmk8GXkdalZpMhmRO5FiV9X1g7rZcUK96NJ5ngk7e2yg93dDwN4x+XOF0IMFklvQiSCgl2IRFCwC5EICnYhEkHBLkQiDLTgpMPQJgX7anNcCikOhyWZbA1/r/I5nl1VGYlkok1HpCZS6LHFFSNksffTSM+2SiQrC5HMJrDihU0+yUv8CUTqXqJcj+g8WXj9ZyM97BCRRK3N5c1WbKm6Z4LjJfDsNY/4UW7w67QSmVeL9GDLjWQqRjImrUT8MP6a6M4uRCIo2IVIBAW7EImgYBciERTsQiTCQHfjASDPwjuMlUqkPU4nvIPbAW8/1M4jO7uRnVFUuK1VD++AViK7pp2hiI/dSIunQsRH5/NaFl6rUomvb7vOE0mKsUukw5M7mh5WNTLnrZrySGJNK1LgrU1qAwLAyGx43DfxXetqJ9LOK6aSZHytYklDXYTVkGKBKxd5J3xdWUSq0Z1diERQsAuRCAp2IRJBwS5EIijYhUgEBbsQiTBQ6c0AGJMgYjXjGmEZzcDrgWWRHA0jEhoAdOe5nORENmyVuXRViNRwK2Q8uaMRSVwp5/zJlcl4k9aEA/ICt0VUPtRr89RWsPAx291InTa+9KhH+hqVi+xZA/Mt0iapxdewyNVBFPipUG/w67FaiNTXqxOJjdSZA4AmacsVuex1ZxciFRTsQiSCgl2IRFCwC5EICnYhEkHBLkQiLCi9mdn9AD4E4JS7/3p/bAzANwHsAnAEwB3ufm4xJ8wtLLE1wGUo9/B7UjGSGNaNZK91IhlU3TVcW6m2w9KKR7KkvMqz71pFLkNVL1MV9To5ZpmLMrXWBWrLPNKMsxBpodQOn88jchJzHQCsys/VjPTY6hJ502p8Tt6ZobZWJLOtW1pLbRZ5cm2iylnGtUj2ckZK0C3qzv7nAG57w9h9AB53990AHu//LIS4glkw2Pv91s++Yfh2AA/0Hz8A4MPL7JcQYpm53L/ZN7v7ZP/xCfQ6ugohrmCWvEHn7o7It/TM7F4zmzCziampqaWeTghxmVxusJ80s60A0P//FPtFd9/n7nvdfe/GjRsv83RCiKVyucH+MIC7+4/vBvDd5XFHCLFSLEZ6exDA+wCMm9mrAD4D4HMAvmVm9wA4CuCOxZzMHWg2w5/4OxH5pESKL7azGj/XPJeMahWepbZmjh+zQZKT8kqkiGIkD6kQKVTZjMhy3ch7dIVkAmYNfrwSVz3hOE9trRYviNiuhfWkVhZZ+0g2X7fFJdEWf8mQkaKYeax1mHFbOdJrqs1ri6LVjbS96oafd6HJ/ZhZE349uxE5esFgd/ePEdMtC80VQlw56Bt0QiSCgl2IRFCwC5EICnYhEkHBLkQiDLbgpAF5KawNdNs8w8csLNd0u1zy6pR58b91Df60vcCz1Lp5WIbqtElDMQDlVkTiKQ5RWyFSfLHNNEAA9aHwc2vP8eKQWYPrNYVIjzjr8LXK14aPWW5wfaqRc3mwY/xcHpHzKiTbLJaF5pF7YH0Nlwcjqhcs0tcv9/Ax80jW2zBxP3b31p1diERQsAuRCAp2IRJBwS5EIijYhUgEBbsQiTBQ6S1GHpFWrEvSsiJah0cKWLYj2WbtYqTw5VxYImmQgpgAkJGMPQAAycgCAKJQAgAKQ9zHZp1kvdW5j6fP8Vqhrz5znNrmnUuOZ2vh51Zo89e5XFhHbetGucw6vI5fxmOVcLbc2tEKnWNruJRnGZ9XjMilseqozWa4mGY9588ry8Mn80iWpe7sQiSCgl2IRFCwC5EICnYhEkHBLkQiDHg33pGTXfJOg+8isiSIbuStKgNPMmlWIkXLnO+adi28I5y1eNJNu86fl4P70Qbf9R0q8XpszUZ4J/nkSVoAGI9873Fqe2b/S9RWsEgij08Hx8/M8Uuu1ZqjtuFxvlO/vrye2rbvHA+O33rz++mcHVcPU1sRPKGoZfy5NfhLhhGiysznfNIwaaO11PZPQohfARTsQiSCgl2IRFCwC5EICnYhEkHBLkQiLKb90/0APgTglLv/en/sswA+AeBiW9ZPu/sjC57NDSD13yySMJJ5WNoqOne/HWm3U2hH6o81uW26Gm79U+YdgZA5T1qZJQkQAFBp8Xnn+DQcPRTulPvQg1+jcw6dmqS2jdt5M87R0W3UVqmHu3ivrXF57fxsWK4DgJcP/wO1vVY7TW2vToaTdS6c+D6dc8utv0ltN7x7N7URZRYAsK7NL5JOmbSoIm2hAKBZCF/7bpE4opb/x58DuC0w/kV339P/t3CgCyFWlQWD3d2fAHB2AL4IIVaQpfzN/kkz229m95sZ/wqTEOKK4HKD/csArgWwB8AkgM+zXzSze81swswmpqbCf08KIVaeywp2dz/p7h137wL4CoCbIr+7z933uvvejRv5Zo8QYmW5rGA3s62X/PgRAM8vjztCiJViMdLbgwDeB2DczF4F8BkA7zOzPQAcwBEAf7yos5nDC0Q3ishXeS2c9dY2PilzLqHFUoOM+QegSkqT5eWI5hKprTdEMtQAYKYVbjUFAPt/8Sq1/c13HwuOv3jwKJ2zfecYtQ2v4bZWxmvQNdvhjL7ZJs/0m5vm0ps3+HocO8/XccPmrcHxV6Z43b2pv/47apuu83lv/SfXUtuatXxba6gUvn5Gh2KZoKTgHWklBSwi2N39Y4Hhry40TwhxZaFv0AmRCAp2IRJBwS5EIijYhUgEBbsQiTDggpMG9v7SirVQKhI5zHlmWzPn2T9z01yWOz/Hpbc13bCtWeKZXLVIhl395Alqe/b5g9T22JMT1Hb0aFiWG9sUzkIDgEpxhNrm+VNDaZ5LXs6KL5I1BICh4TXUdvXmXdS2vspluWY9/ASySGulc2e4pPi1B/83te3ccYDa1m25itrW7wzLg1urXNIdGwtfVzNzXNrUnV2IRFCwC5EICnYhEkHBLkQiKNiFSAQFuxCJMGDpDTALv7+UWzxbp9YIS2yHjvBiGC8ffpHaJqd5gcJT57gfc6fC/dI6Gc++a9e5PDhX4xLP2fNnqK02wyWvuQvhY161eROdU0fkOZ/gmWiRdnRo5uHXeaS4hc5pFyNSXs776c3MXqC2Q6eOBMfLPBkRlQKXSxt13n/t0Fm+Vhtm+DHXzoXXf8t6Xv/hd0+Hr6t2nWTDQXd2IZJBwS5EIijYhUgEBbsQiaBgFyIRBr4b3yEbv1OneV21v/7+D4PjP3zqMJ0zfY7v1M9H2kbNTvGd3TOvHAofr8R3s1Ed57Z5vnubVUp8XqS+XrUTToR48elf0jnFIf6eb23+3GZaXE3odMPtqyrVUTonlgx16tgxaquWuSzQ7JDkpSbftW7PcAVlfHyY2jZWh6jt9Aa+VrVWWB0anTlJ5+zEuuB4qc7bhunOLkQiKNiFSAQFuxCJoGAXIhEU7EIkgoJdiERYTPunHQC+BmAzeu2e9rn7l8xsDMA3AexCrwXUHe7Oe+MAmJ9v4MBLYbns5V+GZS0AeOHnL4XnvMLlpDZJCAGA9gwvrDbX4nJYk2R+VNby+m5DI2uprVjZRm3uXBo6d4YnjIyQJA7P+fHmLCLlVbmsNcJVHmQtYqxGunsbP2B1fAefxnNTMFwPS5EvPf8TOqeb8ySk9swMtXnO68yNFXnmzdyJs8HxqdP8WnzQw9f3WVJzD1jcnb0N4E/d/XoA7wTwJ2Z2PYD7ADzu7rsBPN7/WQhxhbJgsLv7pLv/rP94BsABANsA3A7ggf6vPQDgwyvlpBBi6bypv9nNbBeAGwA8BWCzu0/2TSfQ+5gvhLhCWXSwm9kwgG8D+JS7v+47pe7uQLgCgpnda2YTZjZx7lz4bxMhxMqzqGA3syJ6gf51d3+oP3zSzLb27VsBBMu4uPs+d9/r7nvXr+e9voUQK8uCwW5mhl4/9gPu/oVLTA8DuLv/+G4A311+94QQy8Vist5+B8BdAJ4zs2f6Y58G8DkA3zKzewAcBXDHQgc6d+48/vKbDwVtnZENdF7Tw58IZs88ReecPMSlvKt28syrkRZvuTNPsrLGt+2ic9ZfxWu/bdjEbZ3IS3PyFG/xU26E5atWaZ7O2byey4ONeZ4BNhSp1ZYNh/Uw73Apr13jz8u6XF9rl/lrNn3sp8Hxwi94TbvNo/x5bRjnrbLOnz1ObcMXjlBbFeFac9NnucQ6kYWzESPdyxYOdnf/EXpN2kLcstB8IcSVgb5BJ0QiKNiFSAQFuxCJoGAXIhEU7EIkwkALTl6YreHRJ/cHbePbdtN5haGwm41ZXpBv969tpbZ3vuMmajvb4Il7z/38ueB4vTYZHAeA7jyX106eCLeTAoCswuWwdoNn9JWHw8KJneOFNOebXF4rchUKc0yjAdA+GT5fVuDZX806zyg7c+oEP9f6cPFFADj98kRwvMI7duHte66ntk2j11BbK+NS2f5nwpmbAFDcFM5InO1wSXGoG9bYel9mDaM7uxCJoGAXIhEU7EIkgoJdiERQsAuRCAp2IRJhoNJbx7u40ApnNmVzvFDe9GRYYquM8Qyka7a/hdrqxjOe4FyT2XPD24PjPz/AJbTDrx2kto0VLq00ilxW7Na49NYqh7PD2o1IplwkE61V5uthzqWmVjOs2c1H+sPFWuY1IxlxrcmwJAoA5alwwZRt14UzzQBgbP1OarMCvz+uHeG93q7Zzc/3zLGw3Lt1nC/IcCfsRyEih+rOLkQiKNiFSAQFuxCJoGAXIhEU7EIkwkB34wt5jvWjw0Fbc5q33Jk+Gd6t3FQt0TnVUb4zOrJmDbVt33IttXVqYR8PHeaJMNbgCRyzbZ4kM1LkO7El47v4TYR3z0vGX2ov89ZQWZfXruvW+frnhbD/xTZ/XvPzXJEZqvDiajyNByiVws978whf++Ex3qLq2p1XU1utxpONpud4RtGWC6fD4+M8GWrr5nDyz/4X+WuiO7sQiaBgFyIRFOxCJIKCXYhEULALkQgKdiESYUHpzcx2APgaei2ZHcA+d/+SmX0WwCcATPV/9dPu/kjsWLkZ1uVhaeDlc1PBcQDISa256e40nfPaUS5PbdjDa9DNzPIadIcOhiWSdpPLU1kk52aowP2vF3lGQ4nn/6A5E05OGQI/V7PDpbf5eW7rtsOtpgCgY0QqM16DLmtx+bVZ42tczXmyztxwOMnn7GmevLR5E5dm22Uuzb52lPvfidTe+6cffH9w/Lfe/g46563bwxLgE4/9TzpnMTp7G8CfuvvPzGwEwNNm9mjf9kV3/y+LOIYQYpVZTK+3SQCT/cczZnYAwLaVdkwIsby8qb/ZzWwXgBsAXGyf+kkz229m95sZ/9qREGLVWXSwm9kwgG8D+JS7XwDwZQDXAtiD3p3/82TevWY2YWYTzUbkD1ghxIqyqGA3syJ6gf51d38IANz9pLt33L0L4CsAgrte7r7P3fe6+95SmVdEEUKsLAsGu5kZgK8COODuX7hk/NKWKx8B8PzyuyeEWC4Wsxv/OwDuAvCcmT3TH/s0gI+Z2R705LgjAP54oQONjY3i43f+QdD2wx89Sef94EdhSWPmZOTPgqu59HbmTFhCA4CREf7po4PwvGPHeGYbSNYVAHibn6tQ4M/t+CSXhrISqQs3xyW0UiVctw4Aam2eU1Zp8XvFHGlPFMtR4/laQK0ekTc7XAIsrA/LchvXbefHq/Jr59SRX1JbqczltT3XvY3a7vqjfxEc37iWZ+YVLJw9WK3wVVzMbvyPAIRE36imLoS4stA36IRIBAW7EImgYBciERTsQiSCgl2IRBhowcnhNWW8+8bdQdvxyUN03pM/Dssu8x1eoDA3bps5z6WyTo3LYccPhgtLzl7ghQaLZf5+2nAuk4xVw4U5ASCv84ytodHwSzrf5gUPix4rOMmlsiyLiGVZWPIqtng2XxNcQitXuP/1Os96a5L2T6crvJhjK6LoTs/x17oyzCXMqSO8Ddjpw0eD4yNv4+mN6yLFVhm6swuRCAp2IRJBwS5EIijYhUgEBbsQiaBgFyIRBiq9wXJk1bDk8bZtvNLVH/yzDwXHj0YykDrdGWob3bqD2oqoUdtvjfxGcPzG33sXnbNjLNyTCwC8y5d/ZOM4tXWbXBsaWhvOvGo3uYyTlSLyWpdLZQ3na1UphKXD+fOzdE6zy6W3aoHLa/N17v+5Wvg6mJoMFzEFgIpxCW18xwZqW1vmr+eu626gtur68DHLvF0hQK8d/nrpzi5EIijYhUgEBbsQiaBgFyIRFOxCJIKCXYhEGKj0VioVsXP71qBty7/8KJ1384c+GBw/N8/lNZ8LF+QDgG7GJZ48kom2bThc6HGuwJexUuDH85z72I70eitG+q8VKuF5LePHK/DlQAe8iCJyXgSy6+H7SJZzWcsbPLPNOlxu7JR4pmKrGV6rbodfO2fORopiRvrzdTIuy61by6+RYjfcXyXr8te5033z92nd2YVIBAW7EImgYBciERTsQiSCgl2IRFhwN97MKgCeAFDu//5fuftnzOwtAL4BYAOApwHc5e6k99BFHADZYWxxV1hXnbEWb4Nk6/nObrfAk0wAXrsOWbgtkDnf6e5mfMeaHQ8AOk3WPgnoeGTXuhH2pcBdRDfSPqkZaUNVIOcCgEIeViGaTX6JlCK18KY7/FzDLZ6Qk1l4p75U5Vkm2zfz3f25bJTaCmTnHwA8cl/tePgaKczz5B8neU2RS3FRd/YGgJvd/R3otWe+zczeCeDPAHzR3a8DcA7APYs4lhBilVgw2L3HxbzEYv+fA7gZwF/1xx8A8OEV8VAIsSwstj973u/gegrAowAOATjv/n8/d70KgCekCyFWnUUFu7t33H0PgO0AbgLA+8++ATO718wmzGxiaoq3ShZCrCxvajfe3c8D+HsA7wIwamYXd9W2AzhO5uxz973uvndjpPqKEGJlWTDYzWyjmY32H1cB3ArgAHpBf7GL/N0AvrtSTgohls5iEmG2AnjAzHL03hy+5e7/y8xeBPANM/tPAH4O4KsLHcgdaJI2RKWYJx6WQjrDXDLyNtcg1kTqqjWNSzINkriS5fw908ETP5hMBgClLJIIE0km6ZJ2U60Wl7zceLJOxbgEmBW5dNjKwjJUqRSRKdvcj2okaajY4ck63Tx8YXWYBAyg6Px5jXI1DHXuIrIyv1ZLrfDrGRFtUSASpjl3YsFgd/f9AP6/annufhi9v9+FEP8I0DfohEgEBbsQiaBgFyIRFOxCJIKCXYhEMI9s1S/7ycymABzt/zgO4Er4Sp38eD3y4/X8Y/PjanffGDIMNNhfd2KzCXffuyonlx/yI0E/9DFeiERQsAuRCKsZ7PtW8dyXIj9ej/x4Pb8yfqza3+xCiMGij/FCJMKqBLuZ3WZmL5vZQTO7bzV86PtxxMyeM7NnzGxigOe938xOmdnzl4yNmdmjZvZK//9wT6CV9+OzZna8vybPmFm499by+rHDzP7ezF40sxfM7N/2xwe6JhE/BromZlYxs5+Y2bN9P/5jf/wtZvZUP26+aRZJVwzh7gP9ByBHr6zVNQBKAJ4FcP2g/ej7cgTA+Cqc970AbgTw/CVj/xnAff3H9wH4s1Xy47MA/t2A12MrgBv7j0cA/ALA9YNek4gfA10TAAZguP+4COApAO8E8C0Ad/bH/xuAf/Nmjrsad/abABx098PeKz39DQC3r4Ifq4a7PwHg7BuGb0evcCcwoAKexI+B4+6T7v6z/uMZ9IqjbMOA1yTix0DxHste5HU1gn0bgGOX/LyaxSodwPfM7Gkzu3eVfLjIZnef7D8+AWDzKvrySTPb3/+Yv+J/TlyKme1Cr37CU1jFNXmDH8CA12QlirymvkH3Hne/EcDvA/gTM3vvajsE9N7Z0XsjWg2+DOBa9HoETAL4/KBObGbDAL4N4FPufuFS2yDXJODHwNfEl1DklbEawX4cwI5LfqbFKlcadz/e//8UgO9gdSvvnDSzrQDQ///Uajjh7if7F1oXwFcwoDUxsyJ6AfZ1d3+oPzzwNQn5sVpr0j/3my7yyliNYP8pgN39ncUSgDsBPDxoJ8xsyMxGLj4G8AEAz8dnrSgPo1e4E1jFAp4Xg6vPRzCANTEzQ6+G4QF3/8IlpoGuCfNj0GuyYkVeB7XD+Ibdxg+it9N5CMC/XyUfrkFPCXgWwAuD9APAg+h9HGyh97fXPej1zHscwCsAHgMwtkp+/AWA5wDsRy/Ytg7Aj/eg9xF9P4Bn+v8+OOg1ifgx0DUB8BvoFXHdj94by3+45Jr9CYCDAP4SQPnNHFffoBMiEVLfoBMiGRTsQiSCgl2IRFCwC5EICnYhEkHBLkQiKNiFSAQFuxCJ8H8AlqhGOVqFm0MAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Perturbation Heat Map:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Source image:')\n",
    "imshow(oimg)\n",
    "\n",
    "print('Starting image:')\n",
    "timg, _ = testset[t]\n",
    "timg = torch.unsqueeze(timg, 0).cuda()\n",
    "imshow(timg)\n",
    "\n",
    "print('Adversarial Example:')\n",
    "imshow(adv)\n",
    "\n",
    "print('Perturbation Heat Map:')\n",
    "perturbation_heat_map(oimg,adv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "8c0ccbe2",
   "metadata": {},
   "outputs": [],
   "source": [
    "DSopt = Dt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6e250dc",
   "metadata": {},
   "source": [
    "### 4. RamBoAttack Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01add320",
   "metadata": {},
   "source": [
    "### 4a. RamBoAttack (HSJA) Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "1316559b",
   "metadata": {},
   "outputs": [],
   "source": [
    "i = 2#0,1,2,10,20,50,123 # the sample i-th in the evaluation set\n",
    "query_limit = 50000\n",
    "D = np.zeros(query_limit+2000)\n",
    "nquery = 0\n",
    "o = ID_set[i,1] #oID\n",
    "\n",
    "# 0. select original image\n",
    "oimg, olabel = testset[o]\n",
    "oimg = torch.unsqueeze(oimg, 0).cuda()\n",
    "\n",
    "# 1. select starting image\n",
    "if targeted:\n",
    "    t = ID_set[i,3] #tID, 3 is index acrross dataset - 4 is sample index in a class (not accross dataset)\n",
    "    tlabel = ID_set[i,2]\n",
    "    timg, _ = testset[t]\n",
    "    timg = torch.unsqueeze(timg, 0).cuda()\n",
    "else:\n",
    "    tlabel = None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "4818135d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration: 1, l2 distance 9.4617E+00 - nqry:140\n",
      "iteration: 11, l2 distance 3.2992E+00 - nqry:2821\n",
      "iteration: 21, l2 distance 2.3422E+00 - nqry:7093\n",
      "\n",
      "Break due to slow convergence!\n",
      "\n",
      "Module 1: Finished HSJA\n",
      "\n",
      "Qry# 0 ; l2 distance = 2.339737892150879 ; adv label: 5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/viet/Documents/GitHub/RamBoAttack/ramboattack.py:39: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').\n",
      "  c1 = n // (x.shape[2] * x.shape[3])\n",
      "/home/viet/Documents/GitHub/RamBoAttack/ramboattack.py:40: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').\n",
      "  c2 = (n - c1 * x.shape[2] * x.shape[3])// x.shape[3]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Qry# 1000 ; l2 distance = 1.9404840469360352 ; adv label: 5\n",
      "Qry# 2000 ; l2 distance = 1.7159346342086792 ; adv label: 5\n",
      "Qry# 3000 ; l2 distance = 1.5465682744979858 ; adv label: 5\n",
      "Qry# 4000 ; l2 distance = 1.3929954767227173 ; adv label: 5\n",
      "Qry# 5000 ; l2 distance = 1.3171253204345703 ; adv label: 5\n",
      "Qry# 6000 ; l2 distance = 1.2378469705581665 ; adv label: 5\n",
      "Qry# 7000 ; l2 distance = 1.1951160430908203 ; adv label: 5\n",
      "Qry# 8000 ; l2 distance = 1.1143417358398438 ; adv label: 5\n",
      "Qry# 9000 ; l2 distance = 1.0624281167984009 ; adv label: 5\n",
      "Qry# 10000 ; l2 distance = 1.0075312852859497 ; adv label: 5\n",
      "Qry# 11000 ; l2 distance = 0.973685622215271 ; adv label: 5\n",
      "\n",
      "Break due to slow convergence!\n",
      "\n",
      "Module 2: Finished BlockDescent\n",
      "Targeted attack - Source: 1 and Target: 5 - pred y0: tensor([1], device='cuda:0') - pred yt: tensor([5], device='cuda:0')\n",
      "--------> Found distortion 0.9686\n",
      "==========> Found best distortion 0.9686 in 0.0203 seconds using 17 queries\n",
      "Iteration   1 distortion 0.9684 num_queries 1632\n",
      "Iteration  11 distortion 0.9647 num_queries 10394\n",
      "Iteration  21 distortion 0.9622 num_queries 19609\n",
      "Iteration  31 distortion 0.9596 num_queries 29319\n",
      "break due to query cnt > query limit 31165\n",
      "\n",
      "Adversarial Example Found Successfully: distortion 0.9590 target 5 queries 31165 LS queries 24548 \n",
      "Time: 55.1134 seconds\n",
      "Module 3: Finished SignOPT\n"
     ]
    }
   ],
   "source": [
    "attack_mode = 'RBH' # 'RBH' means RamBoAttack(HSJA) while 'RBS' means RamBoAttack(Sign-OPT) -> see our paper\n",
    "seed = 0\n",
    "query_limit = 50000\n",
    "\n",
    "model_ex = PytorchModel_ex(net, bounds, num_classes,dataset,unnorm)\n",
    "module = RamBoAtt(model_rb,model_ex,testset,seed,targeted,dataset)\n",
    "adv, nqry, Dt = module.hybrid_attack(oimg,olabel,timg,tlabel,query_limit,attack_mode)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "7dd943ad",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Source image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adversarial Example:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Perturbation Heat Map:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Source image:')\n",
    "imshow(oimg)\n",
    "\n",
    "print('Starting image:')\n",
    "timg, _ = testset[t]\n",
    "timg = torch.unsqueeze(timg, 0).cuda()\n",
    "imshow(timg)\n",
    "\n",
    "print('Adversarial Example:')\n",
    "imshow(adv)\n",
    "\n",
    "print('Perturbation Heat Map:')\n",
    "perturbation_heat_map(oimg,adv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "68bb54df",
   "metadata": {},
   "outputs": [],
   "source": [
    "DRambH = Dt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "802fd04e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "- sample ID: 105\n",
      "- ground-truth: 1 \n",
      "- target label: 5 \n",
      "- AdvEx label: 5\n",
      "- L2 dist:0.9589738249778748\n"
     ]
    }
   ],
   "source": [
    "# check prediction of adversarial example\n",
    "alabel = model_rb.predict_label(adv).item()\n",
    "dist = torch.norm(adv-oimg)\n",
    "print(f'- sample ID: {o}\\n- ground-truth: {olabel} \\n- target label: {tlabel} \\n- AdvEx label: {alabel}\\n- L2 dist:{dist}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "80a19e36",
   "metadata": {},
   "source": [
    "### 4b. RamBoAttack (SignOPT) Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "eda3cb9d",
   "metadata": {},
   "outputs": [],
   "source": [
    "i = 2#0,1,2,10,20,50,123 # the sample i-th in the evaluation set\n",
    "query_limit = 50000\n",
    "D = np.zeros(query_limit+2000)\n",
    "nquery = 0\n",
    "o = ID_set[i,1] #oID\n",
    "\n",
    "# 0. select original image\n",
    "oimg, olabel = testset[o]\n",
    "oimg = torch.unsqueeze(oimg, 0).cuda()\n",
    "\n",
    "# 1. select starting image\n",
    "if targeted:\n",
    "    t = ID_set[i,3] #tID, 3 is index acrross dataset - 4 is sample index in a class (not accross dataset)\n",
    "    tlabel = ID_set[i,2]\n",
    "    timg, _ = testset[t]\n",
    "    timg = torch.unsqueeze(timg, 0).cuda()\n",
    "else:\n",
    "    tlabel = None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "6a23aa61",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Targeted attack - Source: 1 and Target: 5 - pred y0: tensor([1], device='cuda:0') - pred yt: tensor([5], device='cuda:0')\n",
      "--------> Found distortion 11.6200\n",
      "==========> Found best distortion 11.6200 in 0.0314 seconds using 22 queries\n",
      "Iteration   1 distortion 9.2015 num_queries 373\n",
      "Iteration  11 distortion 3.9571 num_queries 4240\n",
      "Iteration  21 distortion 3.4252 num_queries 7691\n",
      "\n",
      " break due to slow convergence and group dim = 1!\n",
      "\n",
      "\n",
      "Adversarial Example Found Successfully: distortion 3.4252 target 5 queries 7691 LS queries 3469 \n",
      "Time: 18.0082 seconds\n",
      "Module 1: Finished SignOPT\n",
      "\n",
      "Qry# 0 ; l2 distance = 3.4251983165740967 ; adv label: 5\n",
      "Qry# 1000 ; l2 distance = 2.504671573638916 ; adv label: 5\n",
      "Qry# 2000 ; l2 distance = 2.094240665435791 ; adv label: 5\n",
      "Qry# 3000 ; l2 distance = 1.8161439895629883 ; adv label: 5\n",
      "Qry# 4000 ; l2 distance = 1.6032710075378418 ; adv label: 5\n",
      "Qry# 5000 ; l2 distance = 1.446956753730774 ; adv label: 5\n",
      "Qry# 6000 ; l2 distance = 1.3120777606964111 ; adv label: 5\n",
      "Qry# 7000 ; l2 distance = 1.2575907707214355 ; adv label: 5\n",
      "Qry# 8000 ; l2 distance = 1.2034538984298706 ; adv label: 5\n",
      "Qry# 9000 ; l2 distance = 1.1618778705596924 ; adv label: 5\n",
      "Qry# 10000 ; l2 distance = 1.1220440864562988 ; adv label: 5\n",
      "Qry# 11000 ; l2 distance = 1.0941466093063354 ; adv label: 5\n",
      "\n",
      "Break due to slow convergence!\n",
      "\n",
      "Module 2: Finished BlockDescent\n",
      "Targeted attack - Source: 1 and Target: 5 - pred y0: tensor([1], device='cuda:0') - pred yt: tensor([5], device='cuda:0')\n",
      "--------> Found distortion 1.0672\n",
      "==========> Found best distortion 1.0672 in 0.0248 seconds using 17 queries\n",
      "Iteration   1 distortion 1.0661 num_queries 1145\n",
      "Iteration  11 distortion 1.0574 num_queries 9774\n",
      "Iteration  21 distortion 1.0552 num_queries 18868\n",
      "Iteration  31 distortion 1.0533 num_queries 28420\n",
      "break due to query cnt > query limit 30708\n",
      "\n",
      "Adversarial Example Found Successfully: distortion 1.0530 target 5 queries 30708 LS queries 24091 \n",
      "Time: 54.3417 seconds\n",
      "Module 3: Finished SignOPT\n"
     ]
    }
   ],
   "source": [
    "attack_mode = 'RBS' # 'RBH' means RamBoAttack(HSJA) while 'RBS' means RamBoAttack(Sign-OPT) -> see our paper\n",
    "seed = 0\n",
    "query_limit = 50000\n",
    "module = RamBoAtt(model_rb,model_ex,testset,seed,targeted,dataset)\n",
    "adv, nqry, Dt = module.hybrid_attack(oimg,olabel,timg,tlabel,query_limit,attack_mode)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "ccab5090",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Source image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting image:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adversarial Example:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Perturbation Heat Map:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Source image:')\n",
    "imshow(oimg)\n",
    "\n",
    "print('Starting image:')\n",
    "timg, _ = testset[t]\n",
    "timg = torch.unsqueeze(timg, 0).cuda()\n",
    "imshow(timg)\n",
    "\n",
    "print('Adversarial Example:')\n",
    "imshow(adv)\n",
    "\n",
    "print('Perturbation Heat Map:')\n",
    "perturbation_heat_map(oimg,adv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "0b1490bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "DRambS = Dt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "c677b40c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "- sample ID: 105\n",
      "- ground-truth: 1 \n",
      "- target label: 5 \n",
      "- AdvEx label: 5\n",
      "- L2 dist:1.0529546737670898\n"
     ]
    }
   ],
   "source": [
    "# check prediction of adversarial example\n",
    "alabel = model_rb.predict_label(adv).item()\n",
    "dist = torch.norm(adv-oimg)\n",
    "print(f'- sample ID: {o}\\n- ground-truth: {olabel} \\n- target label: {tlabel} \\n- AdvEx label: {alabel}\\n- L2 dist:{dist}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "059ce735",
   "metadata": {},
   "source": [
    "### 6. Plot Distortion vs Queries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "6ccfcc77",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nq = query_limit\n",
    "q = np.arange(nq)\n",
    "df = pd.DataFrame({'qry':q,'dRambH':np.log(DRambH[:query_limit]),\n",
    "                   'dRambS':np.log(DRambS[:query_limit]),\n",
    "                   'dHsja':np.log(DHsja[:query_limit]),\n",
    "                  'dSopt':np.log(DSopt[:query_limit])})\n",
    "\n",
    "fig_dims = (8, 5)\n",
    "fig, ax = plt.subplots(figsize=fig_dims)\n",
    "f = sns.lineplot(x = 'qry', y = 'dRambH',ax=ax,linewidth = 3, color = 'red',label = 'RamBoAttack (HSJA)',data=df)\n",
    "f = sns.lineplot(x = 'qry', y = 'dRambS',ax=ax,linewidth = 3, color = 'blue',label = 'RamBoAttack (SOPT)',data=df)\n",
    "f = sns.lineplot(x = 'qry', y = 'dHsja',ax=ax,linewidth = 3, color = 'green',label = 'HopSkipJump',data=df)\n",
    "f = sns.lineplot(x = 'qry', y = 'dSopt',ax=ax,linewidth = 3, color = 'orange',label = 'Sign-OPT',data=df)\n",
    "\n",
    "f.axes.set_title(\"Distortion vs Queries (RamBoAttack)\",fontsize=20)\n",
    "f.set_xlabel(\"Queries\",fontsize=18)\n",
    "f.set_ylabel(\"Log10 (L2 dist)\",fontsize=18)\n",
    "f.tick_params(labelsize=15)\n",
    "#plt.legend(labels=['RamBoAttack (HSJA)'])\n",
    "plt.legend(title='Attack', fontsize='14', title_fontsize='18',bbox_to_anchor= (1.5,1))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf1df8a9",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}