├── .gitignore
├── LICENSE
├── README.md
├── keystroke dynamics authentication.ipynb
├── model.h5
└── model.png


/.gitignore:
--------------------------------------------------------------------------------
1 | datasets
2 | .ipynb_checkpoints


--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
 1 | MIT License
 2 | 
 3 | Copyright (c) 2018 Piyush Malhotra
 4 | 
 5 | Permission is hereby granted, free of charge, to any person obtaining a copy
 6 | of this software and associated documentation files (the "Software"), to deal
 7 | in the Software without restriction, including without limitation the rights
 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 | 
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 | 
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # Keystroke Dynamics Using LSTMs
 2 | This repository is a demo for using a LSTM based model to predict whether the person trying to login is authentic or not using keystroke dynamics.
 3 | 
 4 | ## Dependencies
 5 | 1. Keras
 6 | 2. TensorFlow
 7 | 3. Numpy
 8 | 4. Pandas
 9 | 5. Jupyter
10 | 
11 | ## Usage
12 | Run ```jupyter notebook keystroke dynamics authentication.ipynb``` to run the notebook
13 | 
14 | ## Dataset
15 | ```Cs.cmu.edu. (2018). Keystroke Dynamics - Benchmark Data Set. [online] Available at: https://www.cs.cmu.edu/~keystroke/```


--------------------------------------------------------------------------------
/keystroke dynamics authentication.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "markdown",
   5 |    "metadata": {},
   6 |    "source": [
   7 |     "# Keystroke Dynamics Using RNNs\n",
   8 |     "This notebook goes from the process of cleaning and preparing data to train and test a RNN for the job of Keystroke Dynamics. The Authentication Procedure comprises of using the timing of key strike, key hold, key change and so on to reach a conclusion whether the subject under observation is authentic or not."
   9 |    ]
  10 |   },
  11 |   {
  12 |    "cell_type": "markdown",
  13 |    "metadata": {},
  14 |    "source": [
  15 |     "## Data Preparation & Cleaning"
  16 |    ]
  17 |   },
  18 |   {
  19 |    "cell_type": "code",
  20 |    "execution_count": 1,
  21 |    "metadata": {},
  22 |    "outputs": [],
  23 |    "source": [
  24 |     "import pandas as pd\n",
  25 |     "import numpy as np\n",
  26 |     "import matplotlib.pyplot as plt\n",
  27 |     "\n",
  28 |     "%matplotlib inline"
  29 |    ]
  30 |   },
  31 |   {
  32 |    "cell_type": "code",
  33 |    "execution_count": 2,
  34 |    "metadata": {},
  35 |    "outputs": [
  36 |     {
  37 |      "data": {
  38 |       "text/html": [
  39 |        "<div>\n",
  40 |        "<table border=\"1\" class=\"dataframe\">\n",
  41 |        "  <thead>\n",
  42 |        "    <tr style=\"text-align: right;\">\n",
  43 |        "      <th></th>\n",
  44 |        "      <th>subject</th>\n",
  45 |        "      <th>sessionIndex</th>\n",
  46 |        "      <th>rep</th>\n",
  47 |        "      <th>H.period</th>\n",
  48 |        "      <th>DD.period.t</th>\n",
  49 |        "      <th>UD.period.t</th>\n",
  50 |        "      <th>H.t</th>\n",
  51 |        "      <th>DD.t.i</th>\n",
  52 |        "      <th>UD.t.i</th>\n",
  53 |        "      <th>H.i</th>\n",
  54 |        "      <th>...</th>\n",
  55 |        "      <th>H.a</th>\n",
  56 |        "      <th>DD.a.n</th>\n",
  57 |        "      <th>UD.a.n</th>\n",
  58 |        "      <th>H.n</th>\n",
  59 |        "      <th>DD.n.l</th>\n",
  60 |        "      <th>UD.n.l</th>\n",
  61 |        "      <th>H.l</th>\n",
  62 |        "      <th>DD.l.Return</th>\n",
  63 |        "      <th>UD.l.Return</th>\n",
  64 |        "      <th>H.Return</th>\n",
  65 |        "    </tr>\n",
  66 |        "  </thead>\n",
  67 |        "  <tbody>\n",
  68 |        "    <tr>\n",
  69 |        "      <th>0</th>\n",
  70 |        "      <td>s002</td>\n",
  71 |        "      <td>1</td>\n",
  72 |        "      <td>1</td>\n",
  73 |        "      <td>0.1491</td>\n",
  74 |        "      <td>0.3979</td>\n",
  75 |        "      <td>0.2488</td>\n",
  76 |        "      <td>0.1069</td>\n",
  77 |        "      <td>0.1674</td>\n",
  78 |        "      <td>0.0605</td>\n",
  79 |        "      <td>0.1169</td>\n",
  80 |        "      <td>...</td>\n",
  81 |        "      <td>0.1349</td>\n",
  82 |        "      <td>0.1484</td>\n",
  83 |        "      <td>0.0135</td>\n",
  84 |        "      <td>0.0932</td>\n",
  85 |        "      <td>0.3515</td>\n",
  86 |        "      <td>0.2583</td>\n",
  87 |        "      <td>0.1338</td>\n",
  88 |        "      <td>0.3509</td>\n",
  89 |        "      <td>0.2171</td>\n",
  90 |        "      <td>0.0742</td>\n",
  91 |        "    </tr>\n",
  92 |        "    <tr>\n",
  93 |        "      <th>1</th>\n",
  94 |        "      <td>s002</td>\n",
  95 |        "      <td>1</td>\n",
  96 |        "      <td>2</td>\n",
  97 |        "      <td>0.1111</td>\n",
  98 |        "      <td>0.3451</td>\n",
  99 |        "      <td>0.2340</td>\n",
 100 |        "      <td>0.0694</td>\n",
 101 |        "      <td>0.1283</td>\n",
 102 |        "      <td>0.0589</td>\n",
 103 |        "      <td>0.0908</td>\n",
 104 |        "      <td>...</td>\n",
 105 |        "      <td>0.1412</td>\n",
 106 |        "      <td>0.2558</td>\n",
 107 |        "      <td>0.1146</td>\n",
 108 |        "      <td>0.1146</td>\n",
 109 |        "      <td>0.2642</td>\n",
 110 |        "      <td>0.1496</td>\n",
 111 |        "      <td>0.0839</td>\n",
 112 |        "      <td>0.2756</td>\n",
 113 |        "      <td>0.1917</td>\n",
 114 |        "      <td>0.0747</td>\n",
 115 |        "    </tr>\n",
 116 |        "    <tr>\n",
 117 |        "      <th>2</th>\n",
 118 |        "      <td>s002</td>\n",
 119 |        "      <td>1</td>\n",
 120 |        "      <td>3</td>\n",
 121 |        "      <td>0.1328</td>\n",
 122 |        "      <td>0.2072</td>\n",
 123 |        "      <td>0.0744</td>\n",
 124 |        "      <td>0.0731</td>\n",
 125 |        "      <td>0.1291</td>\n",
 126 |        "      <td>0.0560</td>\n",
 127 |        "      <td>0.0821</td>\n",
 128 |        "      <td>...</td>\n",
 129 |        "      <td>0.1621</td>\n",
 130 |        "      <td>0.2332</td>\n",
 131 |        "      <td>0.0711</td>\n",
 132 |        "      <td>0.1172</td>\n",
 133 |        "      <td>0.2705</td>\n",
 134 |        "      <td>0.1533</td>\n",
 135 |        "      <td>0.1085</td>\n",
 136 |        "      <td>0.2847</td>\n",
 137 |        "      <td>0.1762</td>\n",
 138 |        "      <td>0.0945</td>\n",
 139 |        "    </tr>\n",
 140 |        "    <tr>\n",
 141 |        "      <th>3</th>\n",
 142 |        "      <td>s002</td>\n",
 143 |        "      <td>1</td>\n",
 144 |        "      <td>4</td>\n",
 145 |        "      <td>0.1291</td>\n",
 146 |        "      <td>0.2515</td>\n",
 147 |        "      <td>0.1224</td>\n",
 148 |        "      <td>0.1059</td>\n",
 149 |        "      <td>0.2495</td>\n",
 150 |        "      <td>0.1436</td>\n",
 151 |        "      <td>0.1040</td>\n",
 152 |        "      <td>...</td>\n",
 153 |        "      <td>0.1457</td>\n",
 154 |        "      <td>0.1629</td>\n",
 155 |        "      <td>0.0172</td>\n",
 156 |        "      <td>0.0866</td>\n",
 157 |        "      <td>0.2341</td>\n",
 158 |        "      <td>0.1475</td>\n",
 159 |        "      <td>0.0845</td>\n",
 160 |        "      <td>0.3232</td>\n",
 161 |        "      <td>0.2387</td>\n",
 162 |        "      <td>0.0813</td>\n",
 163 |        "    </tr>\n",
 164 |        "    <tr>\n",
 165 |        "      <th>4</th>\n",
 166 |        "      <td>s002</td>\n",
 167 |        "      <td>1</td>\n",
 168 |        "      <td>5</td>\n",
 169 |        "      <td>0.1249</td>\n",
 170 |        "      <td>0.2317</td>\n",
 171 |        "      <td>0.1068</td>\n",
 172 |        "      <td>0.0895</td>\n",
 173 |        "      <td>0.1676</td>\n",
 174 |        "      <td>0.0781</td>\n",
 175 |        "      <td>0.0903</td>\n",
 176 |        "      <td>...</td>\n",
 177 |        "      <td>0.1312</td>\n",
 178 |        "      <td>0.1582</td>\n",
 179 |        "      <td>0.0270</td>\n",
 180 |        "      <td>0.0884</td>\n",
 181 |        "      <td>0.2517</td>\n",
 182 |        "      <td>0.1633</td>\n",
 183 |        "      <td>0.0903</td>\n",
 184 |        "      <td>0.2517</td>\n",
 185 |        "      <td>0.1614</td>\n",
 186 |        "      <td>0.0818</td>\n",
 187 |        "    </tr>\n",
 188 |        "  </tbody>\n",
 189 |        "</table>\n",
 190 |        "<p>5 rows × 34 columns</p>\n",
 191 |        "</div>"
 192 |       ],
 193 |       "text/plain": [
 194 |        "  subject  sessionIndex  rep  H.period  DD.period.t  UD.period.t     H.t  \\\n",
 195 |        "0    s002             1    1    0.1491       0.3979       0.2488  0.1069   \n",
 196 |        "1    s002             1    2    0.1111       0.3451       0.2340  0.0694   \n",
 197 |        "2    s002             1    3    0.1328       0.2072       0.0744  0.0731   \n",
 198 |        "3    s002             1    4    0.1291       0.2515       0.1224  0.1059   \n",
 199 |        "4    s002             1    5    0.1249       0.2317       0.1068  0.0895   \n",
 200 |        "\n",
 201 |        "   DD.t.i  UD.t.i     H.i    ...        H.a  DD.a.n  UD.a.n     H.n  DD.n.l  \\\n",
 202 |        "0  0.1674  0.0605  0.1169    ...     0.1349  0.1484  0.0135  0.0932  0.3515   \n",
 203 |        "1  0.1283  0.0589  0.0908    ...     0.1412  0.2558  0.1146  0.1146  0.2642   \n",
 204 |        "2  0.1291  0.0560  0.0821    ...     0.1621  0.2332  0.0711  0.1172  0.2705   \n",
 205 |        "3  0.2495  0.1436  0.1040    ...     0.1457  0.1629  0.0172  0.0866  0.2341   \n",
 206 |        "4  0.1676  0.0781  0.0903    ...     0.1312  0.1582  0.0270  0.0884  0.2517   \n",
 207 |        "\n",
 208 |        "   UD.n.l     H.l  DD.l.Return  UD.l.Return  H.Return  \n",
 209 |        "0  0.2583  0.1338       0.3509       0.2171    0.0742  \n",
 210 |        "1  0.1496  0.0839       0.2756       0.1917    0.0747  \n",
 211 |        "2  0.1533  0.1085       0.2847       0.1762    0.0945  \n",
 212 |        "3  0.1475  0.0845       0.3232       0.2387    0.0813  \n",
 213 |        "4  0.1633  0.0903       0.2517       0.1614    0.0818  \n",
 214 |        "\n",
 215 |        "[5 rows x 34 columns]"
 216 |       ]
 217 |      },
 218 |      "execution_count": 2,
 219 |      "metadata": {},
 220 |      "output_type": "execute_result"
 221 |     }
 222 |    ],
 223 |    "source": [
 224 |     "df = pd.read_csv('./datasets/data.csv')\n",
 225 |     "\n",
 226 |     "df.head()"
 227 |    ]
 228 |   },
 229 |   {
 230 |    "cell_type": "code",
 231 |    "execution_count": 3,
 232 |    "metadata": {},
 233 |    "outputs": [],
 234 |    "source": [
 235 |     "subjects = df['subject'].unique()"
 236 |    ]
 237 |   },
 238 |   {
 239 |    "cell_type": "code",
 240 |    "execution_count": 4,
 241 |    "metadata": {},
 242 |    "outputs": [],
 243 |    "source": [
 244 |     "subjects_to_int = {subject: i  for i, subject in enumerate(subjects)}\n",
 245 |     "int_to_subjects = {i: subject for i, subject in enumerate(subjects)}"
 246 |    ]
 247 |   },
 248 |   {
 249 |    "cell_type": "code",
 250 |    "execution_count": 5,
 251 |    "metadata": {},
 252 |    "outputs": [],
 253 |    "source": [
 254 |     "df = df.replace(subjects_to_int)"
 255 |    ]
 256 |   },
 257 |   {
 258 |    "cell_type": "code",
 259 |    "execution_count": 6,
 260 |    "metadata": {},
 261 |    "outputs": [
 262 |     {
 263 |      "data": {
 264 |       "text/html": [
 265 |        "<div>\n",
 266 |        "<table border=\"1\" class=\"dataframe\">\n",
 267 |        "  <thead>\n",
 268 |        "    <tr style=\"text-align: right;\">\n",
 269 |        "      <th></th>\n",
 270 |        "      <th>subject</th>\n",
 271 |        "      <th>sessionIndex</th>\n",
 272 |        "      <th>rep</th>\n",
 273 |        "      <th>H.period</th>\n",
 274 |        "      <th>DD.period.t</th>\n",
 275 |        "      <th>UD.period.t</th>\n",
 276 |        "      <th>H.t</th>\n",
 277 |        "      <th>DD.t.i</th>\n",
 278 |        "      <th>UD.t.i</th>\n",
 279 |        "      <th>H.i</th>\n",
 280 |        "      <th>...</th>\n",
 281 |        "      <th>H.a</th>\n",
 282 |        "      <th>DD.a.n</th>\n",
 283 |        "      <th>UD.a.n</th>\n",
 284 |        "      <th>H.n</th>\n",
 285 |        "      <th>DD.n.l</th>\n",
 286 |        "      <th>UD.n.l</th>\n",
 287 |        "      <th>H.l</th>\n",
 288 |        "      <th>DD.l.Return</th>\n",
 289 |        "      <th>UD.l.Return</th>\n",
 290 |        "      <th>H.Return</th>\n",
 291 |        "    </tr>\n",
 292 |        "  </thead>\n",
 293 |        "  <tbody>\n",
 294 |        "    <tr>\n",
 295 |        "      <th>0</th>\n",
 296 |        "      <td>0</td>\n",
 297 |        "      <td>1</td>\n",
 298 |        "      <td>1</td>\n",
 299 |        "      <td>0.1491</td>\n",
 300 |        "      <td>0.3979</td>\n",
 301 |        "      <td>0.2488</td>\n",
 302 |        "      <td>0.1069</td>\n",
 303 |        "      <td>0.1674</td>\n",
 304 |        "      <td>0.0605</td>\n",
 305 |        "      <td>0.1169</td>\n",
 306 |        "      <td>...</td>\n",
 307 |        "      <td>0.1349</td>\n",
 308 |        "      <td>0.1484</td>\n",
 309 |        "      <td>0.0135</td>\n",
 310 |        "      <td>0.0932</td>\n",
 311 |        "      <td>0.3515</td>\n",
 312 |        "      <td>0.2583</td>\n",
 313 |        "      <td>0.1338</td>\n",
 314 |        "      <td>0.3509</td>\n",
 315 |        "      <td>0.2171</td>\n",
 316 |        "      <td>0.0742</td>\n",
 317 |        "    </tr>\n",
 318 |        "    <tr>\n",
 319 |        "      <th>1</th>\n",
 320 |        "      <td>0</td>\n",
 321 |        "      <td>1</td>\n",
 322 |        "      <td>2</td>\n",
 323 |        "      <td>0.1111</td>\n",
 324 |        "      <td>0.3451</td>\n",
 325 |        "      <td>0.2340</td>\n",
 326 |        "      <td>0.0694</td>\n",
 327 |        "      <td>0.1283</td>\n",
 328 |        "      <td>0.0589</td>\n",
 329 |        "      <td>0.0908</td>\n",
 330 |        "      <td>...</td>\n",
 331 |        "      <td>0.1412</td>\n",
 332 |        "      <td>0.2558</td>\n",
 333 |        "      <td>0.1146</td>\n",
 334 |        "      <td>0.1146</td>\n",
 335 |        "      <td>0.2642</td>\n",
 336 |        "      <td>0.1496</td>\n",
 337 |        "      <td>0.0839</td>\n",
 338 |        "      <td>0.2756</td>\n",
 339 |        "      <td>0.1917</td>\n",
 340 |        "      <td>0.0747</td>\n",
 341 |        "    </tr>\n",
 342 |        "    <tr>\n",
 343 |        "      <th>2</th>\n",
 344 |        "      <td>0</td>\n",
 345 |        "      <td>1</td>\n",
 346 |        "      <td>3</td>\n",
 347 |        "      <td>0.1328</td>\n",
 348 |        "      <td>0.2072</td>\n",
 349 |        "      <td>0.0744</td>\n",
 350 |        "      <td>0.0731</td>\n",
 351 |        "      <td>0.1291</td>\n",
 352 |        "      <td>0.0560</td>\n",
 353 |        "      <td>0.0821</td>\n",
 354 |        "      <td>...</td>\n",
 355 |        "      <td>0.1621</td>\n",
 356 |        "      <td>0.2332</td>\n",
 357 |        "      <td>0.0711</td>\n",
 358 |        "      <td>0.1172</td>\n",
 359 |        "      <td>0.2705</td>\n",
 360 |        "      <td>0.1533</td>\n",
 361 |        "      <td>0.1085</td>\n",
 362 |        "      <td>0.2847</td>\n",
 363 |        "      <td>0.1762</td>\n",
 364 |        "      <td>0.0945</td>\n",
 365 |        "    </tr>\n",
 366 |        "    <tr>\n",
 367 |        "      <th>3</th>\n",
 368 |        "      <td>0</td>\n",
 369 |        "      <td>1</td>\n",
 370 |        "      <td>4</td>\n",
 371 |        "      <td>0.1291</td>\n",
 372 |        "      <td>0.2515</td>\n",
 373 |        "      <td>0.1224</td>\n",
 374 |        "      <td>0.1059</td>\n",
 375 |        "      <td>0.2495</td>\n",
 376 |        "      <td>0.1436</td>\n",
 377 |        "      <td>0.1040</td>\n",
 378 |        "      <td>...</td>\n",
 379 |        "      <td>0.1457</td>\n",
 380 |        "      <td>0.1629</td>\n",
 381 |        "      <td>0.0172</td>\n",
 382 |        "      <td>0.0866</td>\n",
 383 |        "      <td>0.2341</td>\n",
 384 |        "      <td>0.1475</td>\n",
 385 |        "      <td>0.0845</td>\n",
 386 |        "      <td>0.3232</td>\n",
 387 |        "      <td>0.2387</td>\n",
 388 |        "      <td>0.0813</td>\n",
 389 |        "    </tr>\n",
 390 |        "    <tr>\n",
 391 |        "      <th>4</th>\n",
 392 |        "      <td>0</td>\n",
 393 |        "      <td>1</td>\n",
 394 |        "      <td>5</td>\n",
 395 |        "      <td>0.1249</td>\n",
 396 |        "      <td>0.2317</td>\n",
 397 |        "      <td>0.1068</td>\n",
 398 |        "      <td>0.0895</td>\n",
 399 |        "      <td>0.1676</td>\n",
 400 |        "      <td>0.0781</td>\n",
 401 |        "      <td>0.0903</td>\n",
 402 |        "      <td>...</td>\n",
 403 |        "      <td>0.1312</td>\n",
 404 |        "      <td>0.1582</td>\n",
 405 |        "      <td>0.0270</td>\n",
 406 |        "      <td>0.0884</td>\n",
 407 |        "      <td>0.2517</td>\n",
 408 |        "      <td>0.1633</td>\n",
 409 |        "      <td>0.0903</td>\n",
 410 |        "      <td>0.2517</td>\n",
 411 |        "      <td>0.1614</td>\n",
 412 |        "      <td>0.0818</td>\n",
 413 |        "    </tr>\n",
 414 |        "  </tbody>\n",
 415 |        "</table>\n",
 416 |        "<p>5 rows × 34 columns</p>\n",
 417 |        "</div>"
 418 |       ],
 419 |       "text/plain": [
 420 |        "   subject  sessionIndex  rep  H.period  DD.period.t  UD.period.t     H.t  \\\n",
 421 |        "0        0             1    1    0.1491       0.3979       0.2488  0.1069   \n",
 422 |        "1        0             1    2    0.1111       0.3451       0.2340  0.0694   \n",
 423 |        "2        0             1    3    0.1328       0.2072       0.0744  0.0731   \n",
 424 |        "3        0             1    4    0.1291       0.2515       0.1224  0.1059   \n",
 425 |        "4        0             1    5    0.1249       0.2317       0.1068  0.0895   \n",
 426 |        "\n",
 427 |        "   DD.t.i  UD.t.i     H.i    ...        H.a  DD.a.n  UD.a.n     H.n  DD.n.l  \\\n",
 428 |        "0  0.1674  0.0605  0.1169    ...     0.1349  0.1484  0.0135  0.0932  0.3515   \n",
 429 |        "1  0.1283  0.0589  0.0908    ...     0.1412  0.2558  0.1146  0.1146  0.2642   \n",
 430 |        "2  0.1291  0.0560  0.0821    ...     0.1621  0.2332  0.0711  0.1172  0.2705   \n",
 431 |        "3  0.2495  0.1436  0.1040    ...     0.1457  0.1629  0.0172  0.0866  0.2341   \n",
 432 |        "4  0.1676  0.0781  0.0903    ...     0.1312  0.1582  0.0270  0.0884  0.2517   \n",
 433 |        "\n",
 434 |        "   UD.n.l     H.l  DD.l.Return  UD.l.Return  H.Return  \n",
 435 |        "0  0.2583  0.1338       0.3509       0.2171    0.0742  \n",
 436 |        "1  0.1496  0.0839       0.2756       0.1917    0.0747  \n",
 437 |        "2  0.1533  0.1085       0.2847       0.1762    0.0945  \n",
 438 |        "3  0.1475  0.0845       0.3232       0.2387    0.0813  \n",
 439 |        "4  0.1633  0.0903       0.2517       0.1614    0.0818  \n",
 440 |        "\n",
 441 |        "[5 rows x 34 columns]"
 442 |       ]
 443 |      },
 444 |      "execution_count": 6,
 445 |      "metadata": {},
 446 |      "output_type": "execute_result"
 447 |     }
 448 |    ],
 449 |    "source": [
 450 |     "df.head()"
 451 |    ]
 452 |   },
 453 |   {
 454 |    "cell_type": "code",
 455 |    "execution_count": 7,
 456 |    "metadata": {},
 457 |    "outputs": [],
 458 |    "source": [
 459 |     "data = df.as_matrix()"
 460 |    ]
 461 |   },
 462 |   {
 463 |    "cell_type": "code",
 464 |    "execution_count": 8,
 465 |    "metadata": {},
 466 |    "outputs": [],
 467 |    "source": [
 468 |     "def generate_positives(data, n_pos_per_subject=150):\n",
 469 |     "    n_subjects = np.unique(data[:, 0]).shape[0]\n",
 470 |     "    poss = []\n",
 471 |     "    for i in range(n_subjects):\n",
 472 |     "        temp_d = data[data[:, 0] == i]\n",
 473 |     "        first_half = temp_d[np.random.choice(400, n_pos_per_subject), 3:-1]\n",
 474 |     "        second_half = temp_d[np.random.choice(400, n_pos_per_subject), 3:-1]\n",
 475 |     "        poss.append(np.hstack([first_half, second_half]))\n",
 476 |     "    return np.vstack(poss)"
 477 |    ]
 478 |   },
 479 |   {
 480 |    "cell_type": "code",
 481 |    "execution_count": 9,
 482 |    "metadata": {},
 483 |    "outputs": [
 484 |     {
 485 |      "data": {
 486 |       "text/plain": [
 487 |        "(7650, 60)"
 488 |       ]
 489 |      },
 490 |      "execution_count": 9,
 491 |      "metadata": {},
 492 |      "output_type": "execute_result"
 493 |     }
 494 |    ],
 495 |    "source": [
 496 |     "poss = generate_positives(data)\n",
 497 |     "poss.shape"
 498 |    ]
 499 |   },
 500 |   {
 501 |    "cell_type": "code",
 502 |    "execution_count": 10,
 503 |    "metadata": {},
 504 |    "outputs": [],
 505 |    "source": [
 506 |     "def generate_negatives(data, n_neg_per_subject=150):\n",
 507 |     "    n_subjects = np.unique(data[:, 0]).shape[0]\n",
 508 |     "    negs = []\n",
 509 |     "    for i in range(n_subjects):\n",
 510 |     "        temp_d = data[data[:, 0] == i]\n",
 511 |     "        temp_not_d = data[data[:, 0] != i]\n",
 512 |     "        first_half = temp_d[np.random.choice(400, n_neg_per_subject), 3:-1]\n",
 513 |     "        second_half = temp_not_d[np.random.choice(400, n_neg_per_subject), 3:-1]\n",
 514 |     "        negs.append(np.hstack([first_half, second_half]))\n",
 515 |     "    return np.vstack(negs)"
 516 |    ]
 517 |   },
 518 |   {
 519 |    "cell_type": "code",
 520 |    "execution_count": 11,
 521 |    "metadata": {},
 522 |    "outputs": [
 523 |     {
 524 |      "data": {
 525 |       "text/plain": [
 526 |        "(7650, 60)"
 527 |       ]
 528 |      },
 529 |      "execution_count": 11,
 530 |      "metadata": {},
 531 |      "output_type": "execute_result"
 532 |     }
 533 |    ],
 534 |    "source": [
 535 |     "negs = generate_negatives(data)\n",
 536 |     "negs.shape"
 537 |    ]
 538 |   },
 539 |   {
 540 |    "cell_type": "code",
 541 |    "execution_count": 12,
 542 |    "metadata": {},
 543 |    "outputs": [
 544 |     {
 545 |      "data": {
 546 |       "text/plain": [
 547 |        "(15300, 1)"
 548 |       ]
 549 |      },
 550 |      "execution_count": 12,
 551 |      "metadata": {},
 552 |      "output_type": "execute_result"
 553 |     }
 554 |    ],
 555 |    "source": [
 556 |     "labels = np.zeros(poss.shape[0] + negs.shape[0])\n",
 557 |     "labels[:poss.shape[0]] = 1\n",
 558 |     "labels = np.expand_dims(labels, axis=1)\n",
 559 |     "labels.shape"
 560 |    ]
 561 |   },
 562 |   {
 563 |    "cell_type": "code",
 564 |    "execution_count": 13,
 565 |    "metadata": {},
 566 |    "outputs": [
 567 |     {
 568 |      "data": {
 569 |       "text/plain": [
 570 |        "array([[ 0.1024,  0.1622,  0.0598, ...,  0.3598,  0.2524,  1.    ],\n",
 571 |        "       [ 0.1074,  0.1525,  0.0451, ...,  0.2532,  0.1466,  1.    ],\n",
 572 |        "       [ 0.0953,  0.1032,  0.0079, ...,  0.2263,  0.1413,  1.    ],\n",
 573 |        "       ..., \n",
 574 |        "       [ 0.0765,  0.1231,  0.0466, ...,  0.192 ,  0.1073,  0.    ],\n",
 575 |        "       [ 0.1031,  0.1049,  0.0018, ...,  0.2313,  0.1252,  0.    ],\n",
 576 |        "       [ 0.1055,  0.1089,  0.0034, ...,  0.2089,  0.1279,  0.    ]])"
 577 |       ]
 578 |      },
 579 |      "execution_count": 13,
 580 |      "metadata": {},
 581 |      "output_type": "execute_result"
 582 |     }
 583 |    ],
 584 |    "source": [
 585 |     "all_data = np.hstack([np.vstack([poss, negs]), labels])\n",
 586 |     "all_data"
 587 |    ]
 588 |   },
 589 |   {
 590 |    "cell_type": "code",
 591 |    "execution_count": 14,
 592 |    "metadata": {},
 593 |    "outputs": [],
 594 |    "source": [
 595 |     "np.random.shuffle(all_data)"
 596 |    ]
 597 |   },
 598 |   {
 599 |    "cell_type": "code",
 600 |    "execution_count": 15,
 601 |    "metadata": {},
 602 |    "outputs": [],
 603 |    "source": [
 604 |     "all_data_t = np.zeros((all_data.shape[0], 15, 4))\n",
 605 |     "\n",
 606 |     "ctr = 0\n",
 607 |     "for i, j in zip(range(0, 30, 2), range(30, 60, 2)):\n",
 608 |     "    all_data_t[:, ctr, :] = np.hstack([all_data[:, i:i+2], all_data[:, j:j+2]])\n",
 609 |     "    ctr += 1"
 610 |    ]
 611 |   },
 612 |   {
 613 |    "cell_type": "code",
 614 |    "execution_count": 16,
 615 |    "metadata": {},
 616 |    "outputs": [],
 617 |    "source": [
 618 |     "X, y = all_data_t, all_data[:, -1]"
 619 |    ]
 620 |   },
 621 |   {
 622 |    "cell_type": "markdown",
 623 |    "metadata": {},
 624 |    "source": [
 625 |     "## Training Phase\n",
 626 |     "This Phase defines parameters of the model, the model itself along with its training to produce a simple RNN that can predict whether a person is the genuine holder of account or not."
 627 |    ]
 628 |   },
 629 |   {
 630 |    "cell_type": "markdown",
 631 |    "metadata": {},
 632 |    "source": [
 633 |     "### Some Params and HyperParams"
 634 |    ]
 635 |   },
 636 |   {
 637 |    "cell_type": "code",
 638 |    "execution_count": 17,
 639 |    "metadata": {},
 640 |    "outputs": [
 641 |     {
 642 |      "name": "stderr",
 643 |      "output_type": "stream",
 644 |      "text": [
 645 |       "Using TensorFlow backend.\n"
 646 |      ]
 647 |     }
 648 |    ],
 649 |    "source": [
 650 |     "from keras.layers import Dense, Activation, Dropout, Input, LSTM, Reshape, Lambda, RepeatVector, Concatenate\n",
 651 |     "from keras.initializers import glorot_uniform\n",
 652 |     "from keras.utils import to_categorical\n",
 653 |     "from keras.models import Model\n",
 654 |     "from keras.optimizers import Adam"
 655 |    ]
 656 |   },
 657 |   {
 658 |    "cell_type": "code",
 659 |    "execution_count": 18,
 660 |    "metadata": {},
 661 |    "outputs": [],
 662 |    "source": [
 663 |     "VALIDATION_SPLIT = 0.1\n",
 664 |     "\n",
 665 |     "INPUT_SHAPE = [None, 2]\n",
 666 |     "\n",
 667 |     "BATCH_SIZE = 32\n",
 668 |     "\n",
 669 |     "EPOCHS = 20"
 670 |    ]
 671 |   },
 672 |   {
 673 |    "cell_type": "code",
 674 |    "execution_count": 19,
 675 |    "metadata": {},
 676 |    "outputs": [],
 677 |    "source": [
 678 |     "def train_dev_split(x, y, val_split=0.1):\n",
 679 |     "    m = x.shape[0]\n",
 680 |     "    val_size = int(0.1 * m)\n",
 681 |     "    return x[:-val_size], y[:, :-val_size, :], x[-val_size:], y[:, -val_size:, :]"
 682 |    ]
 683 |   },
 684 |   {
 685 |    "cell_type": "code",
 686 |    "execution_count": 20,
 687 |    "metadata": {},
 688 |    "outputs": [],
 689 |    "source": [
 690 |     "y = to_categorical(y)\n",
 691 |     "\n",
 692 |     "y_ = np.zeros((15, y.shape[0], y.shape[1]))\n",
 693 |     "\n",
 694 |     "for i in range(15):\n",
 695 |     "    y_[i, :, :] = y"
 696 |    ]
 697 |   },
 698 |   {
 699 |    "cell_type": "code",
 700 |    "execution_count": 21,
 701 |    "metadata": {},
 702 |    "outputs": [
 703 |     {
 704 |      "data": {
 705 |       "text/plain": [
 706 |        "(array([ 0.,  1.]), array([ 0.,  1.]))"
 707 |       ]
 708 |      },
 709 |      "execution_count": 21,
 710 |      "metadata": {},
 711 |      "output_type": "execute_result"
 712 |     }
 713 |    ],
 714 |    "source": [
 715 |     "y_[0, 0, :], y[0]"
 716 |    ]
 717 |   },
 718 |   {
 719 |    "cell_type": "code",
 720 |    "execution_count": 22,
 721 |    "metadata": {},
 722 |    "outputs": [
 723 |     {
 724 |      "data": {
 725 |       "text/plain": [
 726 |        "((13770, 15, 4), (15, 13770, 2), (1530, 15, 4), (15, 1530, 2))"
 727 |       ]
 728 |      },
 729 |      "execution_count": 22,
 730 |      "metadata": {},
 731 |      "output_type": "execute_result"
 732 |     }
 733 |    ],
 734 |    "source": [
 735 |     "x_train, y_train, x_test, y_test = train_dev_split(X, y_)\n",
 736 |     "\n",
 737 |     "x_train.shape, y_train.shape, x_test.shape, y_test.shape"
 738 |    ]
 739 |   },
 740 |   {
 741 |    "cell_type": "code",
 742 |    "execution_count": 23,
 743 |    "metadata": {},
 744 |    "outputs": [],
 745 |    "source": [
 746 |     "n_a = 10\n",
 747 |     "n_out = 2"
 748 |    ]
 749 |   },
 750 |   {
 751 |    "cell_type": "code",
 752 |    "execution_count": 24,
 753 |    "metadata": {},
 754 |    "outputs": [],
 755 |    "source": [
 756 |     "reshapor = Reshape((1, 4))\n",
 757 |     "LSTM_cell = LSTM(n_a, return_state = True)\n",
 758 |     "densor = Dense(n_out, activation='softmax')"
 759 |    ]
 760 |   },
 761 |   {
 762 |    "cell_type": "code",
 763 |    "execution_count": 25,
 764 |    "metadata": {},
 765 |    "outputs": [],
 766 |    "source": [
 767 |     "def keystroke_model(Tx, n_in, n_a, n_out):\n",
 768 |     "    X = Input(shape=(Tx, n_in))\n",
 769 |     "    \n",
 770 |     "    a0 = Input(shape=(n_a,), name='a0')\n",
 771 |     "    c0 = Input(shape=(n_a,), name='c0')\n",
 772 |     "    a = a0\n",
 773 |     "    c = c0\n",
 774 |     "    \n",
 775 |     "    outputs = []\n",
 776 |     "\n",
 777 |     "    for t in range(Tx):\n",
 778 |     "         \n",
 779 |     "        x = Lambda(lambda x: X[:, t, :])(X)\n",
 780 |     "        x = reshapor(x)\n",
 781 |     "\n",
 782 |     "        a, _, c = LSTM_cell(x, initial_state=[a, c])\n",
 783 |     "\n",
 784 |     "        out = densor(a)\n",
 785 |     "\n",
 786 |     "        outputs.append(out)\n",
 787 |     "\n",
 788 |     "    model = Model(inputs=[X, a0, c0], outputs=outputs)\n",
 789 |     "    \n",
 790 |     "    return model"
 791 |    ]
 792 |   },
 793 |   {
 794 |    "cell_type": "code",
 795 |    "execution_count": 26,
 796 |    "metadata": {},
 797 |    "outputs": [],
 798 |    "source": [
 799 |     "model = keystroke_model(15, 4, n_a, n_out)"
 800 |    ]
 801 |   },
 802 |   {
 803 |    "cell_type": "code",
 804 |    "execution_count": 27,
 805 |    "metadata": {},
 806 |    "outputs": [],
 807 |    "source": [
 808 |     "opt = Adam(lr=0.01, beta_1=0.9, beta_2=0.999, decay=0.01)\n",
 809 |     "\n",
 810 |     "model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])"
 811 |    ]
 812 |   },
 813 |   {
 814 |    "cell_type": "code",
 815 |    "execution_count": 28,
 816 |    "metadata": {},
 817 |    "outputs": [
 818 |     {
 819 |      "name": "stdout",
 820 |      "output_type": "stream",
 821 |      "text": [
 822 |       "__________________________________________________________________________________________________\n",
 823 |       "Layer (type)                    Output Shape         Param #     Connected to                     \n",
 824 |       "==================================================================================================\n",
 825 |       "input_1 (InputLayer)            (None, 15, 4)        0                                            \n",
 826 |       "__________________________________________________________________________________________________\n",
 827 |       "lambda_1 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 828 |       "__________________________________________________________________________________________________\n",
 829 |       "reshape_1 (Reshape)             (None, 1, 4)         0           lambda_1[0][0]                   \n",
 830 |       "                                                                 lambda_2[0][0]                   \n",
 831 |       "                                                                 lambda_3[0][0]                   \n",
 832 |       "                                                                 lambda_4[0][0]                   \n",
 833 |       "                                                                 lambda_5[0][0]                   \n",
 834 |       "                                                                 lambda_6[0][0]                   \n",
 835 |       "                                                                 lambda_7[0][0]                   \n",
 836 |       "                                                                 lambda_8[0][0]                   \n",
 837 |       "                                                                 lambda_9[0][0]                   \n",
 838 |       "                                                                 lambda_10[0][0]                  \n",
 839 |       "                                                                 lambda_11[0][0]                  \n",
 840 |       "                                                                 lambda_12[0][0]                  \n",
 841 |       "                                                                 lambda_13[0][0]                  \n",
 842 |       "                                                                 lambda_14[0][0]                  \n",
 843 |       "                                                                 lambda_15[0][0]                  \n",
 844 |       "__________________________________________________________________________________________________\n",
 845 |       "a0 (InputLayer)                 (None, 10)           0                                            \n",
 846 |       "__________________________________________________________________________________________________\n",
 847 |       "c0 (InputLayer)                 (None, 10)           0                                            \n",
 848 |       "__________________________________________________________________________________________________\n",
 849 |       "lambda_2 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 850 |       "__________________________________________________________________________________________________\n",
 851 |       "lstm_1 (LSTM)                   [(None, 10), (None,  600         reshape_1[0][0]                  \n",
 852 |       "                                                                 a0[0][0]                         \n",
 853 |       "                                                                 c0[0][0]                         \n",
 854 |       "                                                                 reshape_1[1][0]                  \n",
 855 |       "                                                                 lstm_1[0][0]                     \n",
 856 |       "                                                                 lstm_1[0][2]                     \n",
 857 |       "                                                                 reshape_1[2][0]                  \n",
 858 |       "                                                                 lstm_1[1][0]                     \n",
 859 |       "                                                                 lstm_1[1][2]                     \n",
 860 |       "                                                                 reshape_1[3][0]                  \n",
 861 |       "                                                                 lstm_1[2][0]                     \n",
 862 |       "                                                                 lstm_1[2][2]                     \n",
 863 |       "                                                                 reshape_1[4][0]                  \n",
 864 |       "                                                                 lstm_1[3][0]                     \n",
 865 |       "                                                                 lstm_1[3][2]                     \n",
 866 |       "                                                                 reshape_1[5][0]                  \n",
 867 |       "                                                                 lstm_1[4][0]                     \n",
 868 |       "                                                                 lstm_1[4][2]                     \n",
 869 |       "                                                                 reshape_1[6][0]                  \n",
 870 |       "                                                                 lstm_1[5][0]                     \n",
 871 |       "                                                                 lstm_1[5][2]                     \n",
 872 |       "                                                                 reshape_1[7][0]                  \n",
 873 |       "                                                                 lstm_1[6][0]                     \n",
 874 |       "                                                                 lstm_1[6][2]                     \n",
 875 |       "                                                                 reshape_1[8][0]                  \n",
 876 |       "                                                                 lstm_1[7][0]                     \n",
 877 |       "                                                                 lstm_1[7][2]                     \n",
 878 |       "                                                                 reshape_1[9][0]                  \n",
 879 |       "                                                                 lstm_1[8][0]                     \n",
 880 |       "                                                                 lstm_1[8][2]                     \n",
 881 |       "                                                                 reshape_1[10][0]                 \n",
 882 |       "                                                                 lstm_1[9][0]                     \n",
 883 |       "                                                                 lstm_1[9][2]                     \n",
 884 |       "                                                                 reshape_1[11][0]                 \n",
 885 |       "                                                                 lstm_1[10][0]                    \n",
 886 |       "                                                                 lstm_1[10][2]                    \n",
 887 |       "                                                                 reshape_1[12][0]                 \n",
 888 |       "                                                                 lstm_1[11][0]                    \n",
 889 |       "                                                                 lstm_1[11][2]                    \n",
 890 |       "                                                                 reshape_1[13][0]                 \n",
 891 |       "                                                                 lstm_1[12][0]                    \n",
 892 |       "                                                                 lstm_1[12][2]                    \n",
 893 |       "                                                                 reshape_1[14][0]                 \n",
 894 |       "                                                                 lstm_1[13][0]                    \n",
 895 |       "                                                                 lstm_1[13][2]                    \n",
 896 |       "__________________________________________________________________________________________________\n",
 897 |       "lambda_3 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 898 |       "__________________________________________________________________________________________________\n",
 899 |       "lambda_4 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 900 |       "__________________________________________________________________________________________________\n",
 901 |       "lambda_5 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 902 |       "__________________________________________________________________________________________________\n",
 903 |       "lambda_6 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 904 |       "__________________________________________________________________________________________________\n",
 905 |       "lambda_7 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 906 |       "__________________________________________________________________________________________________\n",
 907 |       "lambda_8 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 908 |       "__________________________________________________________________________________________________\n",
 909 |       "lambda_9 (Lambda)               (None, 4)            0           input_1[0][0]                    \n",
 910 |       "__________________________________________________________________________________________________\n",
 911 |       "lambda_10 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 912 |       "__________________________________________________________________________________________________\n",
 913 |       "lambda_11 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 914 |       "__________________________________________________________________________________________________\n",
 915 |       "lambda_12 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 916 |       "__________________________________________________________________________________________________\n",
 917 |       "lambda_13 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 918 |       "__________________________________________________________________________________________________\n",
 919 |       "lambda_14 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 920 |       "__________________________________________________________________________________________________\n",
 921 |       "lambda_15 (Lambda)              (None, 4)            0           input_1[0][0]                    \n",
 922 |       "__________________________________________________________________________________________________\n",
 923 |       "dense_1 (Dense)                 (None, 2)            22          lstm_1[0][0]                     \n",
 924 |       "                                                                 lstm_1[1][0]                     \n",
 925 |       "                                                                 lstm_1[2][0]                     \n",
 926 |       "                                                                 lstm_1[3][0]                     \n",
 927 |       "                                                                 lstm_1[4][0]                     \n",
 928 |       "                                                                 lstm_1[5][0]                     \n",
 929 |       "                                                                 lstm_1[6][0]                     \n",
 930 |       "                                                                 lstm_1[7][0]                     \n",
 931 |       "                                                                 lstm_1[8][0]                     \n",
 932 |       "                                                                 lstm_1[9][0]                     \n",
 933 |       "                                                                 lstm_1[10][0]                    \n",
 934 |       "                                                                 lstm_1[11][0]                    \n",
 935 |       "                                                                 lstm_1[12][0]                    \n",
 936 |       "                                                                 lstm_1[13][0]                    \n",
 937 |       "                                                                 lstm_1[14][0]                    \n",
 938 |       "==================================================================================================\n",
 939 |       "Total params: 622\n",
 940 |       "Trainable params: 622\n",
 941 |       "Non-trainable params: 0\n",
 942 |       "__________________________________________________________________________________________________\n"
 943 |      ]
 944 |     }
 945 |    ],
 946 |    "source": [
 947 |     "model.summary()"
 948 |    ]
 949 |   },
 950 |   {
 951 |    "cell_type": "code",
 952 |    "execution_count": 29,
 953 |    "metadata": {},
 954 |    "outputs": [],
 955 |    "source": [
 956 |     "m = x_train.shape[0]\n",
 957 |     "a0 = np.zeros((m, n_a))\n",
 958 |     "c0 = np.zeros((m, n_a))"
 959 |    ]
 960 |   },
 961 |   {
 962 |    "cell_type": "code",
 963 |    "execution_count": 31,
 964 |    "metadata": {},
 965 |    "outputs": [
 966 |     {
 967 |      "name": "stdout",
 968 |      "output_type": "stream",
 969 |      "text": [
 970 |       "Epoch 1/10\n",
 971 |       "13770/13770 [==============================] - 56s 4ms/step - loss: 10.3848 - dense_1_loss_1: 0.6944 - dense_1_loss_2: 0.6916 - dense_1_loss_3: 0.6912 - dense_1_loss_4: 0.6914 - dense_1_loss_5: 0.6906 - dense_1_loss_6: 0.6920 - dense_1_loss_7: 0.6923 - dense_1_loss_8: 0.6927 - dense_1_loss_9: 0.6925 - dense_1_loss_10: 0.6931 - dense_1_loss_11: 0.6929 - dense_1_loss_12: 0.6928 - dense_1_loss_13: 0.6927 - dense_1_loss_14: 0.6927 - dense_1_loss_15: 0.6918 - dense_1_acc_1: 0.4810 - dense_1_acc_2: 0.5246 - dense_1_acc_3: 0.5251 - dense_1_acc_4: 0.5248 - dense_1_acc_5: 0.5301 - dense_1_acc_6: 0.5130 - dense_1_acc_7: 0.5113 - dense_1_acc_8: 0.5106 - dense_1_acc_9: 0.5076 - dense_1_acc_10: 0.5081 - dense_1_acc_11: 0.5081 - dense_1_acc_12: 0.5078 - dense_1_acc_13: 0.5093 - dense_1_acc_14: 0.5089 - dense_1_acc_15: 0.5086: 11s - loss: 10.3932 - dense_1_loss_1: 0.6946 - dense_1_loss_2: 0.6922 - dense_1_loss_3: 0.6920 - dense_1_loss_4: 0.6922 - dense_1_loss_5: 0.6916 - dense_1_loss_6: 0.6925 - dense_1_loss_7: 0.6928 - dense_1_loss_8: 0.6930 - dense_1_loss_9: 0.6932 - dense_1_loss_10: 0.6935 - dense_1_loss_11: 0.6933 - dense_1_loss_12: 0.6933 - dense_1_loss_13: 0.6932 - dense_1_loss_14: 0.6932 - dense_1_loss_15: 0.6927 - dense_1_acc_1: 0.4784 - dense_1_acc_2: 0.5163 - dense_1_acc_3: 0.5151 - dense_1_acc_4: 0.5158 - dense_1_acc_5: 0.5212 - dense_1_acc_6: 0.5150 - dense_1_acc_7: 0.5135 - dense_1_acc_8: 0.5130 - dense_1_acc_9: 0.5055 - dense_1_acc_10: 0.5064 - dense_1_acc_11: 0.50\n",
 972 |       "Epoch 2/10\n",
 973 |       "13770/13770 [==============================] - 49s 4ms/step - loss: 9.6077 - dense_1_loss_1: 0.6934 - dense_1_loss_2: 0.6853 - dense_1_loss_3: 0.6792 - dense_1_loss_4: 0.6721 - dense_1_loss_5: 0.6626 - dense_1_loss_6: 0.6462 - dense_1_loss_7: 0.6392 - dense_1_loss_8: 0.6265 - dense_1_loss_9: 0.6190 - dense_1_loss_10: 0.6149 - dense_1_loss_11: 0.6108 - dense_1_loss_12: 0.6105 - dense_1_loss_13: 0.6103 - dense_1_loss_14: 0.6139 - dense_1_loss_15: 0.6240 - dense_1_acc_1: 0.4833 - dense_1_acc_2: 0.5985 - dense_1_acc_3: 0.6121 - dense_1_acc_4: 0.6129 - dense_1_acc_5: 0.6158 - dense_1_acc_6: 0.6251 - dense_1_acc_7: 0.6381 - dense_1_acc_8: 0.6356 - dense_1_acc_9: 0.6411 - dense_1_acc_10: 0.6469 - dense_1_acc_11: 0.6493 - dense_1_acc_12: 0.6477 - dense_1_acc_13: 0.6486 - dense_1_acc_14: 0.6434 - dense_1_acc_15: 0.6433\n",
 974 |       "Epoch 3/10\n",
 975 |       "13770/13770 [==============================] - 49s 4ms/step - loss: 9.0376 - dense_1_loss_1: 0.6912 - dense_1_loss_2: 0.6796 - dense_1_loss_3: 0.6677 - dense_1_loss_4: 0.6541 - dense_1_loss_5: 0.6386 - dense_1_loss_6: 0.6075 - dense_1_loss_7: 0.5959 - dense_1_loss_8: 0.5776 - dense_1_loss_9: 0.5702 - dense_1_loss_10: 0.5629 - dense_1_loss_11: 0.5564 - dense_1_loss_12: 0.5562 - dense_1_loss_13: 0.5541 - dense_1_loss_14: 0.5566 - dense_1_loss_15: 0.5688 - dense_1_acc_1: 0.5065 - dense_1_acc_2: 0.6169 - dense_1_acc_3: 0.6268 - dense_1_acc_4: 0.6336 - dense_1_acc_5: 0.6393 - dense_1_acc_6: 0.6777 - dense_1_acc_7: 0.7076 - dense_1_acc_8: 0.7195 - dense_1_acc_9: 0.7211 - dense_1_acc_10: 0.7277 - dense_1_acc_11: 0.7283 - dense_1_acc_12: 0.7275 - dense_1_acc_13: 0.7284 - dense_1_acc_14: 0.7253 - dense_1_acc_15: 0.7248\n",
 976 |       "Epoch 4/10\n",
 977 |       "13770/13770 [==============================] - 53s 4ms/step - loss: 8.7896 - dense_1_loss_1: 0.6898 - dense_1_loss_2: 0.6765 - dense_1_loss_3: 0.6621 - dense_1_loss_4: 0.6466 - dense_1_loss_5: 0.6297 - dense_1_loss_6: 0.5935 - dense_1_loss_7: 0.5781 - dense_1_loss_8: 0.5577 - dense_1_loss_9: 0.5499 - dense_1_loss_10: 0.5406 - dense_1_loss_11: 0.5328 - dense_1_loss_12: 0.5324 - dense_1_loss_13: 0.5293 - dense_1_loss_14: 0.5303 - dense_1_loss_15: 0.5402 - dense_1_acc_1: 0.5075 - dense_1_acc_2: 0.6200 - dense_1_acc_3: 0.6325 - dense_1_acc_4: 0.6382 - dense_1_acc_5: 0.6402 - dense_1_acc_6: 0.6963 - dense_1_acc_7: 0.7275 - dense_1_acc_8: 0.7453 - dense_1_acc_9: 0.7449 - dense_1_acc_10: 0.7474 - dense_1_acc_11: 0.7522 - dense_1_acc_12: 0.7502 - dense_1_acc_13: 0.7530 - dense_1_acc_14: 0.7525 - dense_1_acc_15: 0.7527\n",
 978 |       "Epoch 5/10\n",
 979 |       "13770/13770 [==============================] - 50s 4ms/step - loss: 8.6522 - dense_1_loss_1: 0.6887 - dense_1_loss_2: 0.6739 - dense_1_loss_3: 0.6579 - dense_1_loss_4: 0.6412 - dense_1_loss_5: 0.6236 - dense_1_loss_6: 0.5845 - dense_1_loss_7: 0.5674 - dense_1_loss_8: 0.5466 - dense_1_loss_9: 0.5389 - dense_1_loss_10: 0.5292 - dense_1_loss_11: 0.5208 - dense_1_loss_12: 0.5201 - dense_1_loss_13: 0.5168 - dense_1_loss_14: 0.5172 - dense_1_loss_15: 0.5254 - dense_1_acc_1: 0.5094 - dense_1_acc_2: 0.6211 - dense_1_acc_3: 0.6336 - dense_1_acc_4: 0.6452 - dense_1_acc_5: 0.6438 - dense_1_acc_6: 0.7056 - dense_1_acc_7: 0.7328 - dense_1_acc_8: 0.7508 - dense_1_acc_9: 0.7524 - dense_1_acc_10: 0.7554 - dense_1_acc_11: 0.7601 - dense_1_acc_12: 0.7588 - dense_1_acc_13: 0.7614 - dense_1_acc_14: 0.7630 - dense_1_acc_15: 0.7628\n",
 980 |       "Epoch 6/10\n",
 981 |       "13770/13770 [==============================] - 50s 4ms/step - loss: 8.5593 - dense_1_loss_1: 0.6877 - dense_1_loss_2: 0.6717 - dense_1_loss_3: 0.6544 - dense_1_loss_4: 0.6365 - dense_1_loss_5: 0.6186 - dense_1_loss_6: 0.5780 - dense_1_loss_7: 0.5602 - dense_1_loss_8: 0.5397 - dense_1_loss_9: 0.5321 - dense_1_loss_10: 0.5218 - dense_1_loss_11: 0.5131 - dense_1_loss_12: 0.5124 - dense_1_loss_13: 0.5086 - dense_1_loss_14: 0.5085 - dense_1_loss_15: 0.5161 - dense_1_acc_1: 0.5081 - dense_1_acc_2: 0.6282 - dense_1_acc_3: 0.6383 - dense_1_acc_4: 0.6476 - dense_1_acc_5: 0.6470 - dense_1_acc_6: 0.7078 - dense_1_acc_7: 0.7378 - dense_1_acc_8: 0.7574 - dense_1_acc_9: 0.7574 - dense_1_acc_10: 0.7619 - dense_1_acc_11: 0.7644 - dense_1_acc_12: 0.7639 - dense_1_acc_13: 0.7658 - dense_1_acc_14: 0.7679 - dense_1_acc_15: 0.7676\n",
 982 |       "Epoch 7/10\n",
 983 |       "13770/13770 [==============================] - 51s 4ms/step - loss: 8.5034 - dense_1_loss_1: 0.6868 - dense_1_loss_2: 0.6698 - dense_1_loss_3: 0.6514 - dense_1_loss_4: 0.6328 - dense_1_loss_5: 0.6147 - dense_1_loss_6: 0.5735 - dense_1_loss_7: 0.5555 - dense_1_loss_8: 0.5358 - dense_1_loss_9: 0.5283 - dense_1_loss_10: 0.5176 - dense_1_loss_11: 0.5087 - dense_1_loss_12: 0.5081 - dense_1_loss_13: 0.5042 - dense_1_loss_14: 0.5044 - dense_1_loss_15: 0.5118 - dense_1_acc_1: 0.5094 - dense_1_acc_2: 0.6304 - dense_1_acc_3: 0.6410 - dense_1_acc_4: 0.6476 - dense_1_acc_5: 0.6470 - dense_1_acc_6: 0.7105 - dense_1_acc_7: 0.7394 - dense_1_acc_8: 0.7588 - dense_1_acc_9: 0.7577 - dense_1_acc_10: 0.7635 - dense_1_acc_11: 0.7676 - dense_1_acc_12: 0.7670 - dense_1_acc_13: 0.7698 - dense_1_acc_14: 0.7710 - dense_1_acc_15: 0.7702\n",
 984 |       "Epoch 8/10\n",
 985 |       "13770/13770 [==============================] - 51s 4ms/step - loss: 8.4593 - dense_1_loss_1: 0.6861 - dense_1_loss_2: 0.6684 - dense_1_loss_3: 0.6493 - dense_1_loss_4: 0.6298 - dense_1_loss_5: 0.6115 - dense_1_loss_6: 0.5701 - dense_1_loss_7: 0.5525 - dense_1_loss_8: 0.5334 - dense_1_loss_9: 0.5259 - dense_1_loss_10: 0.5143 - dense_1_loss_11: 0.5050 - dense_1_loss_12: 0.5044 - dense_1_loss_13: 0.5003 - dense_1_loss_14: 0.5003 - dense_1_loss_15: 0.5079 - dense_1_acc_1: 0.5147 - dense_1_acc_2: 0.6299 - dense_1_acc_3: 0.6415 - dense_1_acc_4: 0.6495 - dense_1_acc_5: 0.6468 - dense_1_acc_6: 0.7113 - dense_1_acc_7: 0.7410 - dense_1_acc_8: 0.7611 - dense_1_acc_9: 0.7605 - dense_1_acc_10: 0.7646 - dense_1_acc_11: 0.7709 - dense_1_acc_12: 0.7683 - dense_1_acc_13: 0.7714 - dense_1_acc_14: 0.7745 - dense_1_acc_15: 0.7747\n",
 986 |       "Epoch 9/10\n",
 987 |       "13770/13770 [==============================] - 51s 4ms/step - loss: 8.4227 - dense_1_loss_1: 0.6855 - dense_1_loss_2: 0.6675 - dense_1_loss_3: 0.6478 - dense_1_loss_4: 0.6282 - dense_1_loss_5: 0.6105 - dense_1_loss_6: 0.5681 - dense_1_loss_7: 0.5491 - dense_1_loss_8: 0.5302 - dense_1_loss_9: 0.5228 - dense_1_loss_10: 0.5106 - dense_1_loss_11: 0.5012 - dense_1_loss_12: 0.5010 - dense_1_loss_13: 0.4971 - dense_1_loss_14: 0.4975 - dense_1_loss_15: 0.5058 - dense_1_acc_1: 0.5182 - dense_1_acc_2: 0.6318 - dense_1_acc_3: 0.6430 - dense_1_acc_4: 0.6506 - dense_1_acc_5: 0.6499 - dense_1_acc_6: 0.7138 - dense_1_acc_7: 0.7463 - dense_1_acc_8: 0.7618 - dense_1_acc_9: 0.7622 - dense_1_acc_10: 0.7650 - dense_1_acc_11: 0.7714 - dense_1_acc_12: 0.7697 - dense_1_acc_13: 0.7744 - dense_1_acc_14: 0.7765 - dense_1_acc_15: 0.7749\n",
 988 |       "Epoch 10/10\n"
 989 |      ]
 990 |     },
 991 |     {
 992 |      "name": "stdout",
 993 |      "output_type": "stream",
 994 |      "text": [
 995 |       "13770/13770 [==============================] - 50s 4ms/step - loss: 8.3825 - dense_1_loss_1: 0.6850 - dense_1_loss_2: 0.6665 - dense_1_loss_3: 0.6463 - dense_1_loss_4: 0.6262 - dense_1_loss_5: 0.6081 - dense_1_loss_6: 0.5651 - dense_1_loss_7: 0.5465 - dense_1_loss_8: 0.5276 - dense_1_loss_9: 0.5200 - dense_1_loss_10: 0.5072 - dense_1_loss_11: 0.4977 - dense_1_loss_12: 0.4974 - dense_1_loss_13: 0.4934 - dense_1_loss_14: 0.4936 - dense_1_loss_15: 0.5019 - dense_1_acc_1: 0.5252 - dense_1_acc_2: 0.6338 - dense_1_acc_3: 0.6433 - dense_1_acc_4: 0.6529 - dense_1_acc_5: 0.6486 - dense_1_acc_6: 0.7179 - dense_1_acc_7: 0.7445 - dense_1_acc_8: 0.7640 - dense_1_acc_9: 0.7635 - dense_1_acc_10: 0.7683 - dense_1_acc_11: 0.7759 - dense_1_acc_12: 0.7744 - dense_1_acc_13: 0.7805 - dense_1_acc_14: 0.7815 - dense_1_acc_15: 0.7796\n"
 996 |      ]
 997 |     },
 998 |     {
 999 |      "data": {
1000 |       "text/plain": [
1001 |        "<keras.callbacks.History at 0x18a8d1009b0>"
1002 |       ]
1003 |      },
1004 |      "execution_count": 31,
1005 |      "metadata": {},
1006 |      "output_type": "execute_result"
1007 |     }
1008 |    ],
1009 |    "source": [
1010 |     "model.fit([x_train, a0, c0], list(y_train), epochs=10)"
1011 |    ]
1012 |   },
1013 |   {
1014 |    "cell_type": "code",
1015 |    "execution_count": 32,
1016 |    "metadata": {},
1017 |    "outputs": [],
1018 |    "source": [
1019 |     "m_t = x_test.shape[0]\n",
1020 |     "a0_t = np.zeros((m_t, n_a))\n",
1021 |     "c0_t = np.zeros((m_t, n_a))"
1022 |    ]
1023 |   },
1024 |   {
1025 |    "cell_type": "code",
1026 |    "execution_count": 34,
1027 |    "metadata": {},
1028 |    "outputs": [
1029 |     {
1030 |      "name": "stdout",
1031 |      "output_type": "stream",
1032 |      "text": [
1033 |       "1530/1530 [==============================] - 3s 2ms/step\n"
1034 |      ]
1035 |     },
1036 |     {
1037 |      "data": {
1038 |       "text/plain": [
1039 |        "[8.1394441193225333,\n",
1040 |        " 0.68649219843297227,\n",
1041 |        " 0.66842293131585218,\n",
1042 |        " 0.6474493969499675,\n",
1043 |        " 0.62604353209726171,\n",
1044 |        " 0.60595984139473613,\n",
1045 |        " 0.55427271182241,\n",
1046 |        " 0.53408347352657448,\n",
1047 |        " 0.51291458419725011,\n",
1048 |        " 0.50179924154593269,\n",
1049 |        " 0.48426854883144105,\n",
1050 |        " 0.47200996883554397,\n",
1051 |        " 0.467594226980521,\n",
1052 |        " 0.45899881054373348,\n",
1053 |        " 0.45738684663585588,\n",
1054 |        " 0.46174781665303349,\n",
1055 |        " 0.5294117646279678,\n",
1056 |        " 0.63333333372290612,\n",
1057 |        " 0.65163398677227546,\n",
1058 |        " 0.65947712457257934,\n",
1059 |        " 0.67450980423322693,\n",
1060 |        " 0.73464052264207325,\n",
1061 |        " 0.76143790880839035,\n",
1062 |        " 0.76928104551789023,\n",
1063 |        " 0.78169934663897245,\n",
1064 |        " 0.78888888912263255,\n",
1065 |        " 0.79607843160629277,\n",
1066 |        " 0.79411764674716523,\n",
1067 |        " 0.79411764674716523,\n",
1068 |        " 0.80196078400206716,\n",
1069 |        " 0.80326797354455093]"
1070 |       ]
1071 |      },
1072 |      "execution_count": 34,
1073 |      "metadata": {},
1074 |      "output_type": "execute_result"
1075 |     }
1076 |    ],
1077 |    "source": [
1078 |     "model.evaluate([x_test, a0_t, c0_t], list(y_test))"
1079 |    ]
1080 |   },
1081 |   {
1082 |    "cell_type": "code",
1083 |    "execution_count": 36,
1084 |    "metadata": {},
1085 |    "outputs": [],
1086 |    "source": [
1087 |     "from keras.utils import plot_model\n",
1088 |     "plot_model(model, to_file='model.png')"
1089 |    ]
1090 |   },
1091 |   {
1092 |    "cell_type": "markdown",
1093 |    "metadata": {},
1094 |    "source": [
1095 |     "![](model.png)"
1096 |    ]
1097 |   },
1098 |   {
1099 |    "cell_type": "code",
1100 |    "execution_count": 45,
1101 |    "metadata": {},
1102 |    "outputs": [],
1103 |    "source": [
1104 |     "model.save_weights('model.h5')"
1105 |    ]
1106 |   }
1107 |  ],
1108 |  "metadata": {
1109 |   "kernelspec": {
1110 |    "display_name": "Python 3",
1111 |    "language": "python",
1112 |    "name": "python3"
1113 |   },
1114 |   "language_info": {
1115 |    "codemirror_mode": {
1116 |     "name": "ipython",
1117 |     "version": 3
1118 |    },
1119 |    "file_extension": ".py",
1120 |    "mimetype": "text/x-python",
1121 |    "name": "python",
1122 |    "nbconvert_exporter": "python",
1123 |    "pygments_lexer": "ipython3",
1124 |    "version": "3.5.1"
1125 |   }
1126 |  },
1127 |  "nbformat": 4,
1128 |  "nbformat_minor": 2
1129 | }
1130 | 


--------------------------------------------------------------------------------
/model.h5:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/piyush2896/Keystroke-Dynamics-Using-LSTMs/bcb1a361e8fad94ac45daa13e48b1b42c68984b2/model.h5


--------------------------------------------------------------------------------
/model.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/piyush2896/Keystroke-Dynamics-Using-LSTMs/bcb1a361e8fad94ac45daa13e48b1b42c68984b2/model.png


--------------------------------------------------------------------------------