do_test_success_stochastic_e5_bat_fb_all.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "from get_mmap import get_mmap\n",
    "import random\n",
    "\n",
    "# Setup the necessary paths and parameters\n",
    "fmap = 'e5_all_bat_fb_2mhz_n500_beta.raw'\n",
    "data_title = 'Stochastic A6 BAT FB'\n",
    "path_data = 'results/'\n",
    "name_data = 'a6_bat_fb_stochastic_f17_all_dlinear_n1000r_p1_slr_g1000_r10.pkl'\n",
    "rand_iter = 10\n",
    "total_values = 2**16\n",
    "pmul = 2**4\n",
    "n_profile = 1000*pmul # ensure that: n_profile + n_attack*nr_attack_groups < nr_trials\n",
    "n_attack = 150\n",
    "n_profile_per_group = 100\n",
    "nr_traces_vec = [1, 2, 5, 10, 20, 50, 100]\n",
    "\n",
    "random.seed(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load file\n",
    "print('Mapping data...')\n",
    "mmap_data, metadata = get_mmap(fmap)\n",
    "print('Done mapping data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "print('Extracting values corresponding to mapped data')\n",
    "D_all_bytes = mmap_data['data'][0]['B'][1:3, :].T\n",
    "D_all = np.bitwise_xor(np.left_shift(D_all_bytes[:, 0].astype(np.uint16), 8), D_all_bytes[:, 1].astype(np.uint16))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Selecting traces for profile and attack...')\n",
    "nr_trials = metadata['nr_trials']\n",
    "total_trials_per_value = nr_trials // total_values\n",
    "idx_group = np.arange(1, total_trials_per_value + 1)\n",
    "\n",
    "idx_profile_group = np.where(idx_group % 4 == 0)[0]\n",
    "total_profile_trials_per_group = len(idx_profile_group)\n",
    "idx_attack_group = np.union1d(np.union1d(np.where(idx_group % 4 == 1)[0], np.where(idx_group % 4 == 2)[0]), \n",
    "                              np.where(idx_group % 4 == 3)[0])\n",
    "\n",
    "total_profile = total_values * total_profile_trials_per_group\n",
    "idx_profile_all = np.zeros(total_profile, dtype=int)\n",
    "\n",
    "si = np.argsort(D_all, axis=0)\n",
    "for k in range(total_values):\n",
    "    idx_val = si[total_trials_per_value * k: total_trials_per_value * (k + 1)]\n",
    "    idx_profile_all[total_profile_trials_per_group * k: total_profile_trials_per_group * (k + 1)] = idx_val[idx_profile_group]\n",
    "\n",
    "idx_profile_sel = np.random.choice(idx_profile_all, n_profile, replace=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Select base for coefficients\n",
    "base = 'F17'\n",
    "\n",
    "# Select values for discriminant evaluation\n",
    "V_discriminant = np.arange(total_values)\n",
    "\n",
    "# Select number of values for V_profile and V_attack\n",
    "ng_profile = [2**16]\n",
    "ng_attack = [2**8]\n",
    "nr_eval = 3\n",
    "\n",
    "# Set up attack/result cells\n",
    "nr_attacks = 5\n",
    "results = np.full((len(ng_attack), len(ng_profile), nr_eval, nr_attacks), None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run attacks\n",
    "for r in range(nr_eval):\n",
    "    for k in range(len(ng_attack)):\n",
    "        # Select groups for V_attack\n",
    "        if ng_attack[k] == total_values:\n",
    "            V_attack = V_discriminant  # full guessing entropy\n",
    "        else:\n",
    "            V_attack = np.random.randint(0, total_values, ng_attack[k])\n",
    "\n",
    "        for i in range(len(ng_profile)):\n",
    "            print(f'Running attacks for (k,i,r)=({k},{i},{r})')\n",
    "\n",
    "            # Select groups for V_profile\n",
    "            if ng_profile[i] == total_values:\n",
    "                V_profile = np.arange(total_values)\n",
    "            else:\n",
    "                V_profile = np.random.randint(0, total_values, ng_profile[i])\n",
    "\n",
    "            # Run attack for selection, 20ppc\n",
    "            atype = 'selection'\n",
    "            aparams = {\n",
    "                'signal': 'bnorm',\n",
    "                'sel': '20ppc',\n",
    "                'p1': 40,\n",
    "                'p2': 0.90\n",
    "            }\n",
    "            discriminant = 'linearfast'\n",
    "            eparams = {'save_signals': 1}\n",
    "            results[k, i, r, 0] = run_stochastic_attack_generic_mmap(\n",
    "                mmap_data, D_all, idx_profile_sel,\n",
    "                mmap_data, D_all, idx_attack_group,\n",
    "                V_profile, V_attack, V_discriminant,\n",
    "                base, atype, aparams, discriminant,\n",
    "                rand_iter, nr_traces_vec, eparams\n",
    "            )\n",
    "\n",
    "            # Run attack for S-PCA (aka Master PCA)\n",
    "            atype = 'templatepca'\n",
    "            aparams = {\n",
    "                'pca_threshold': 0.95,\n",
    "                'pca_alternate': 0,\n",
    "                'pca_dimensions': 10\n",
    "            }\n",
    "            discriminant = 'linearfast'\n",
    "            eparams = {}\n",
    "            results[k, i, r, 1] = run_stochastic_attack_generic_mmap(\n",
    "                mmap_data, D_all, idx_profile_sel,\n",
    "                mmap_data, D_all, idx_attack_group,\n",
    "                V_profile, V_attack, V_discriminant,\n",
    "                base, atype, aparams, discriminant,\n",
    "                rand_iter, nr_traces_vec, eparams\n",
    "            )\n",
    "\n",
    "            # Run attack for T-PCA (aka Slave PCA)\n",
    "            atype = 'pca'\n",
    "            aparams = {\n",
    "                'idx_traces': idx_profile_group,\n",
    "                'pca_threshold': 0.95,\n",
    "                'pca_alternate': 0,\n",
    "                'pca_dimensions': 10,\n",
    "                'cov_from_sel': 0\n",
    "            }\n",
    "            discriminant = 'linearfast'\n",
    "            eparams = {}\n",
    "            results[k, i, r, 2] = run_stochastic_attack_generic_mmap(\n",
    "                mmap_data, D_all, idx_profile_sel,\n",
    "                mmap_data, D_all, idx_attack_group,\n",
    "                V_profile, V_attack, V_discriminant,\n",
    "                base, atype, aparams, discriminant,\n",
    "                rand_iter, nr_traces_vec, eparams\n",
    "            )\n",
    "\n",
    "            # Run attack for S-LDA (aka Master LDA)\n",
    "            atype = 'templatelda'\n",
    "            aparams = {\n",
    "                'lda_threshold': 0.95,\n",
    "                'lda_dimensions': 10\n",
    "            }\n",
    "            discriminant = 'linearnocov'\n",
    "            eparams = {}\n",
    "            results[k, i, r, 3] = run_stochastic_attack_generic_mmap(\n",
    "                mmap_data, D_all, idx_profile_sel,\n",
    "                mmap_data, D_all, idx_attack_group,\n",
    "                V_profile, V_attack, V_discriminant,\n",
    "                base, atype, aparams, discriminant,\n",
    "                rand_iter, nr_traces_vec, eparams\n",
    "            )\n",
    "\n",
    "            # Run attack for T-LDA (aka Slave LDA)\n",
    "            atype = 'lda'\n",
    "            aparams = {\n",
    "                'idx_traces': idx_profile_group,\n",
    "                'lda_threshold': 0.95,\n",
    "                'lda_dimensions': 10,\n",
    "                'cov_from_sel': 0\n",
    "            }\n",
    "            discriminant = 'linearfastnocov'\n",
    "            eparams = {}\n",
    "            results[k, i, r, 4] = run_stochastic_attack_generic_mmap(\n",
    "                mmap_data, D_all, idx_profile_sel,\n",
    "                mmap_data, D_all, idx_attack_group,\n",
    "                V_profile, V_attack, V_discriminant,\n",
    "                base, atype, aparams, discriminant,\n",
    "                rand_iter, nr_traces_vec, eparams\n",
    "            )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pickle\n",
    "\n",
    "# Save data to disk so you can prevent wasting time\n",
    "print('All done, saving data...\\n')\n",
    "os.makedirs(path_data, exist_ok=True)\n",
    "\n",
    "with open(os.path.join(path_data, name_data), 'wb') as f:\n",
    "    pickle.dump(results, f)"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}