4 Hosts as bags of processes over time.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "96246e84-d4c7-40d7-b3db-69f67c687ace",
   "metadata": {},
   "source": [
    "# Embedding hosts as distributions on point clouds of process instances\n",
    "\n",
    "An interesting insight on representing objects with vectors is that they enable _hierarchical_ vector space embeddings.\n",
    "Indeed, if one has a hierarchy of objects containing each other,\n",
    "then a vector representation of the innermost objects can be lifted back up through a sequence of Wasserstein embeddings of the container layers.\n",
    "\n",
    "Concretely, we have process instances, attached to command lines.\n",
    "On the one hand, command lines (and thus process instances) _contain_ a set of tokens.\n",
    "On the other hand, processes run on hosts at a specific time;\n",
    "said differently, hosts, over a certain time interval, _contain_ a set of process instances.\n",
    "So, since we have a vector space embedding of process instances (command lines really),\n",
    "we can get a vector representation of hosts as distributions on a point cloud of process instances.\n",
    "<a id=\"interpretation\"></a>\n",
    "Thus hosts, in a time interval, would accrete into clusters when they run similar process workloads.\n",
    "\n",
    "---\n",
    "\n",
    "Goals of this notebook:\n",
    "\n",
    "1. Compute a vector representation of hosts in a sequence of time intervals.\n",
    "2. Appraise these vectors in a flat data map.\n",
    "3. Derive a naïve visualization of the behaviour of a single host over time.\n",
    "\n",
    "---\n",
    "\n",
    "**Attention**: results from [notebook 2](Command%20lines%20-%20Wasserstein%20embedding.ipynb) are required for what follows, so run that notebook first."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "09cb17d1-23ae-46a5-a869-5bdde42465da",
   "metadata": {},
   "outputs": [],
   "source": [
    "import datamapplot as dmp\n",
    "from fast_hdbscan import HDBSCAN\n",
    "import glasbey\n",
    "from IPython.display import Video\n",
    "from matplotlib import colormaps\n",
    "from matplotlib.animation import FuncAnimation\n",
    "from matplotlib.colors import ListedColormap\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy.signal\n",
    "import scipy.sparse\n",
    "from tqdm.auto import trange, tqdm\n",
    "import umap\n",
    "import vectorizers as vz\n",
    "import vectorizers.transformers as vzt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6525294e-7217-471b-bea1-8ae397f4b58f",
   "metadata": {},
   "source": [
    "Gather our processes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "999a37e4-d93e-4ba2-8b16-4ae0e1d46bc2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>pid_hash</th>\n",
       "      <th>parent_pid_hash</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>hostname</th>\n",
       "      <th>name</th>\n",
       "      <th>cmdline</th>\n",
       "      <th>ptree</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3F8784056EB55BB295DCAB8C9344254A</td>\n",
       "      <td>BA7EC59E059F8E7B9F75D3BD0D3E85FE</td>\n",
       "      <td>2023-11-05 00:00:00.314092-07:00</td>\n",
       "      <td>ACME-HH-YIM</td>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\wsqmcons.exe\"</td>\n",
       "      <td>=wsqmcons.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>B64CB859912B447E83D8C8E619ABCF93</td>\n",
       "      <td>50F07B711DB04C89FF3922CD15C2647B</td>\n",
       "      <td>2023-11-05 00:00:00.377725-07:00</td>\n",
       "      <td>ACME-HH-UWI</td>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\wsqmcons.exe\"</td>\n",
       "      <td>=wsqmcons.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>921D6E35B58C5CA571A55D96C16A5D6A</td>\n",
       "      <td>981581F0C573975A7D576645C5B731F0</td>\n",
       "      <td>2023-11-05 00:00:00.436069-07:00</td>\n",
       "      <td>ACME-HH-HGC</td>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\wsqmcons.exe\"</td>\n",
       "      <td>=wsqmcons.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>22E2B0D56B1F636B09CBDC86CCDBBD97</td>\n",
       "      <td>483B551ED4FD3BF742CBBD9B343E8117</td>\n",
       "      <td>2023-11-05 00:00:00.586099-07:00</td>\n",
       "      <td>ACME-WS-AZU</td>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\wsqmcons.exe\"</td>\n",
       "      <td>=wsqmcons.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1F1B598005B90EB45FBF77E74DDFCB8</td>\n",
       "      <td>803022DFF3341FF443B9D49E08D82558</td>\n",
       "      <td>2023-11-05 00:00:00.731733-07:00</td>\n",
       "      <td>ACME-HH-AKA</td>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\wsqmcons.exe\"</td>\n",
       "      <td>=wsqmcons.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136701</th>\n",
       "      <td>11D2905584ADEBDB004263B08D3706A1</td>\n",
       "      <td>5B1453B37F079EC6AA910E820C5B6BF1</td>\n",
       "      <td>2023-11-20 15:54:14.616846-08:00</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>usoclient.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\usoclient.exe\" startscan</td>\n",
       "      <td>=usoclient.exe-&gt;svchost.exe-&gt;services.exe-&gt;win...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136702</th>\n",
       "      <td>D8DC51881EDB5C6F7D6239E3DE1A5D93</td>\n",
       "      <td>E283B625D417A1C9584B2D6C9D0C364F</td>\n",
       "      <td>2023-11-20 15:54:14.638486-08:00</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>mousocoreworker.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\mousocoreworker.exe\" -emb...</td>\n",
       "      <td>=mousocoreworker.exe-&gt;svchost.exe-&gt;services.ex...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136703</th>\n",
       "      <td>521490D4B45E5BDC4BD2CA7081BC06A7</td>\n",
       "      <td>6A37E544F4D9E91016071993B3DC4103</td>\n",
       "      <td>2023-11-20 15:55:23.333011-08:00</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "      <td>usoclient.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\usoclient.exe\" startscan</td>\n",
       "      <td>=usoclient.exe-&gt;svchost.exe-&gt;services.exe-&gt;win...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136704</th>\n",
       "      <td>D2211D916C3414FDFBD2663BB9051C84</td>\n",
       "      <td>2C2DB3363D7539099C7869C6BFEEED6B</td>\n",
       "      <td>2023-11-20 15:55:23.349693-08:00</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "      <td>mousocoreworker.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\mousocoreworker.exe\" -emb...</td>\n",
       "      <td>=mousocoreworker.exe-&gt;svchost.exe-&gt;services.ex...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136705</th>\n",
       "      <td>73917C5EF988165232D5F62B384F829E</td>\n",
       "      <td>5B1453B37F079EC6AA910E820C5B6BF1</td>\n",
       "      <td>2023-11-20 15:57:56.983102-08:00</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>rundll32.exe</td>\n",
       "      <td>\"c:\\windows\\system32\\rundll32.exe\" c:\\windows\\...</td>\n",
       "      <td>=rundll32.exe-&gt;svchost.exe-&gt;services.exe-&gt;wini...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>136706 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                pid_hash                   parent_pid_hash  \\\n",
       "0       3F8784056EB55BB295DCAB8C9344254A  BA7EC59E059F8E7B9F75D3BD0D3E85FE   \n",
       "1       B64CB859912B447E83D8C8E619ABCF93  50F07B711DB04C89FF3922CD15C2647B   \n",
       "2       921D6E35B58C5CA571A55D96C16A5D6A  981581F0C573975A7D576645C5B731F0   \n",
       "3       22E2B0D56B1F636B09CBDC86CCDBBD97  483B551ED4FD3BF742CBBD9B343E8117   \n",
       "4       F1F1B598005B90EB45FBF77E74DDFCB8  803022DFF3341FF443B9D49E08D82558   \n",
       "...                                  ...                               ...   \n",
       "136701  11D2905584ADEBDB004263B08D3706A1  5B1453B37F079EC6AA910E820C5B6BF1   \n",
       "136702  D8DC51881EDB5C6F7D6239E3DE1A5D93  E283B625D417A1C9584B2D6C9D0C364F   \n",
       "136703  521490D4B45E5BDC4BD2CA7081BC06A7  6A37E544F4D9E91016071993B3DC4103   \n",
       "136704  D2211D916C3414FDFBD2663BB9051C84  2C2DB3363D7539099C7869C6BFEEED6B   \n",
       "136705  73917C5EF988165232D5F62B384F829E  5B1453B37F079EC6AA910E820C5B6BF1   \n",
       "\n",
       "                              timestamp     hostname                 name  \\\n",
       "0      2023-11-05 00:00:00.314092-07:00  ACME-HH-YIM         wsqmcons.exe   \n",
       "1      2023-11-05 00:00:00.377725-07:00  ACME-HH-UWI         wsqmcons.exe   \n",
       "2      2023-11-05 00:00:00.436069-07:00  ACME-HH-HGC         wsqmcons.exe   \n",
       "3      2023-11-05 00:00:00.586099-07:00  ACME-WS-AZU         wsqmcons.exe   \n",
       "4      2023-11-05 00:00:00.731733-07:00  ACME-HH-AKA         wsqmcons.exe   \n",
       "...                                 ...          ...                  ...   \n",
       "136701 2023-11-20 15:54:14.616846-08:00  ACME-HH-IKA        usoclient.exe   \n",
       "136702 2023-11-20 15:54:14.638486-08:00  ACME-HH-IKA  mousocoreworker.exe   \n",
       "136703 2023-11-20 15:55:23.333011-08:00  ACME-HH-AZH        usoclient.exe   \n",
       "136704 2023-11-20 15:55:23.349693-08:00  ACME-HH-AZH  mousocoreworker.exe   \n",
       "136705 2023-11-20 15:57:56.983102-08:00  ACME-HH-IKA         rundll32.exe   \n",
       "\n",
       "                                                  cmdline  \\\n",
       "0                      \"c:\\windows\\system32\\wsqmcons.exe\"   \n",
       "1                      \"c:\\windows\\system32\\wsqmcons.exe\"   \n",
       "2                      \"c:\\windows\\system32\\wsqmcons.exe\"   \n",
       "3                      \"c:\\windows\\system32\\wsqmcons.exe\"   \n",
       "4                      \"c:\\windows\\system32\\wsqmcons.exe\"   \n",
       "...                                                   ...   \n",
       "136701      \"c:\\windows\\system32\\usoclient.exe\" startscan   \n",
       "136702  \"c:\\windows\\system32\\mousocoreworker.exe\" -emb...   \n",
       "136703      \"c:\\windows\\system32\\usoclient.exe\" startscan   \n",
       "136704  \"c:\\windows\\system32\\mousocoreworker.exe\" -emb...   \n",
       "136705  \"c:\\windows\\system32\\rundll32.exe\" c:\\windows\\...   \n",
       "\n",
       "                                                    ptree  \n",
       "0       =wsqmcons.exe->svchost.exe->services.exe->wini...  \n",
       "1       =wsqmcons.exe->svchost.exe->services.exe->wini...  \n",
       "2       =wsqmcons.exe->svchost.exe->services.exe->wini...  \n",
       "3       =wsqmcons.exe->svchost.exe->services.exe->wini...  \n",
       "4       =wsqmcons.exe->svchost.exe->services.exe->wini...  \n",
       "...                                                   ...  \n",
       "136701  =usoclient.exe->svchost.exe->services.exe->win...  \n",
       "136702  =mousocoreworker.exe->svchost.exe->services.ex...  \n",
       "136703  =usoclient.exe->svchost.exe->services.exe->win...  \n",
       "136704  =mousocoreworker.exe->svchost.exe->services.ex...  \n",
       "136705  =rundll32.exe->svchost.exe->services.exe->wini...  \n",
       "\n",
       "[136706 rows x 7 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "processes = pd.read_parquet(\"process_filtered.parquet\")\n",
    "processes"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a079882-792b-413a-8a0b-fcd5dbc21e43",
   "metadata": {},
   "source": [
    "Get the better command line vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "d5880465-eae8-4fb1-b2c0-f2bdf5f12214",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30994, 128)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cmdlines_vec = np.load(\"cmdlines_wasserstein.npz\")[\"cmdlines\"]\n",
    "cmdlines_vec.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4effe28d-c689-4788-9eac-a3535fcd9a23",
   "metadata": {},
   "source": [
    "So our command line vectors were derived with _uniqued_ command lines.\n",
    "We want to take into account _all_ process instances here, so let's tie the index to each command line vector back to the process instance record."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "690fd41a-34fe-4301-9b04-9f7c728ca701",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>pid_hash</th>\n",
       "      <th>hostname</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>index_vec</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>3F8784056EB55BB295DCAB8C9344254A</td>\n",
       "      <td>ACME-HH-YIM</td>\n",
       "      <td>2023-11-05 02:00:00.314092-05:00</td>\n",
       "      <td>30162</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>B64CB859912B447E83D8C8E619ABCF93</td>\n",
       "      <td>ACME-HH-UWI</td>\n",
       "      <td>2023-11-05 02:00:00.377725-05:00</td>\n",
       "      <td>30162</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>921D6E35B58C5CA571A55D96C16A5D6A</td>\n",
       "      <td>ACME-HH-HGC</td>\n",
       "      <td>2023-11-05 02:00:00.436069-05:00</td>\n",
       "      <td>30162</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>22E2B0D56B1F636B09CBDC86CCDBBD97</td>\n",
       "      <td>ACME-WS-AZU</td>\n",
       "      <td>2023-11-05 02:00:00.586099-05:00</td>\n",
       "      <td>30162</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>wsqmcons.exe</td>\n",
       "      <td>F1F1B598005B90EB45FBF77E74DDFCB8</td>\n",
       "      <td>ACME-HH-AKA</td>\n",
       "      <td>2023-11-05 02:00:00.731733-05:00</td>\n",
       "      <td>30162</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136701</th>\n",
       "      <td>usoclient.exe</td>\n",
       "      <td>11D2905584ADEBDB004263B08D3706A1</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>2023-11-20 18:54:14.616846-05:00</td>\n",
       "      <td>30029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136702</th>\n",
       "      <td>mousocoreworker.exe</td>\n",
       "      <td>D8DC51881EDB5C6F7D6239E3DE1A5D93</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>2023-11-20 18:54:14.638486-05:00</td>\n",
       "      <td>28614</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136703</th>\n",
       "      <td>usoclient.exe</td>\n",
       "      <td>521490D4B45E5BDC4BD2CA7081BC06A7</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "      <td>2023-11-20 18:55:23.333011-05:00</td>\n",
       "      <td>30029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136704</th>\n",
       "      <td>mousocoreworker.exe</td>\n",
       "      <td>D2211D916C3414FDFBD2663BB9051C84</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "      <td>2023-11-20 18:55:23.349693-05:00</td>\n",
       "      <td>28614</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136705</th>\n",
       "      <td>rundll32.exe</td>\n",
       "      <td>73917C5EF988165232D5F62B384F829E</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "      <td>2023-11-20 18:57:56.983102-05:00</td>\n",
       "      <td>28874</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>136706 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       name                          pid_hash     hostname  \\\n",
       "0              wsqmcons.exe  3F8784056EB55BB295DCAB8C9344254A  ACME-HH-YIM   \n",
       "1              wsqmcons.exe  B64CB859912B447E83D8C8E619ABCF93  ACME-HH-UWI   \n",
       "2              wsqmcons.exe  921D6E35B58C5CA571A55D96C16A5D6A  ACME-HH-HGC   \n",
       "3              wsqmcons.exe  22E2B0D56B1F636B09CBDC86CCDBBD97  ACME-WS-AZU   \n",
       "4              wsqmcons.exe  F1F1B598005B90EB45FBF77E74DDFCB8  ACME-HH-AKA   \n",
       "...                     ...                               ...          ...   \n",
       "136701        usoclient.exe  11D2905584ADEBDB004263B08D3706A1  ACME-HH-IKA   \n",
       "136702  mousocoreworker.exe  D8DC51881EDB5C6F7D6239E3DE1A5D93  ACME-HH-IKA   \n",
       "136703        usoclient.exe  521490D4B45E5BDC4BD2CA7081BC06A7  ACME-HH-AZH   \n",
       "136704  mousocoreworker.exe  D2211D916C3414FDFBD2663BB9051C84  ACME-HH-AZH   \n",
       "136705         rundll32.exe  73917C5EF988165232D5F62B384F829E  ACME-HH-IKA   \n",
       "\n",
       "                              timestamp  index_vec  \n",
       "0      2023-11-05 02:00:00.314092-05:00      30162  \n",
       "1      2023-11-05 02:00:00.377725-05:00      30162  \n",
       "2      2023-11-05 02:00:00.436069-05:00      30162  \n",
       "3      2023-11-05 02:00:00.586099-05:00      30162  \n",
       "4      2023-11-05 02:00:00.731733-05:00      30162  \n",
       "...                                 ...        ...  \n",
       "136701 2023-11-20 18:54:14.616846-05:00      30029  \n",
       "136702 2023-11-20 18:54:14.638486-05:00      28614  \n",
       "136703 2023-11-20 18:55:23.333011-05:00      30029  \n",
       "136704 2023-11-20 18:55:23.349693-05:00      28614  \n",
       "136705 2023-11-20 18:57:56.983102-05:00      28874  \n",
       "\n",
       "[136706 rows x 5 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with np.load(\"cmdlines_unique.npz\") as store_cmdlines_unique:\n",
    "    i_unique = store_cmdlines_unique[\"i_unique\"]\n",
    "    i_deunique = store_cmdlines_unique[\"i_deunique\"]\n",
    "processes_indexed = processes[[\"name\", \"pid_hash\", \"hostname\"]].copy()\n",
    "processes_indexed[\"timestamp\"] = processes[\"timestamp\"].dt.tz_convert(\"-0500\")\n",
    "processes_indexed[\"index_vec\"] = i_deunique\n",
    "processes_indexed"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fdf07352-f754-459d-a7d7-dad098e8b979",
   "metadata": {},
   "source": [
    "And now, count instances of each command line by host and by time interval.\n",
    "First group process instances by host,\n",
    "then `resample` the process instance timestamps in 3-hour bins.\n",
    "We also ensure to reject all process instances that might formally be outside of the data capture period &mdash;\n",
    "we seem to have some instances timestamped prior to November 5 2023, midnight EST.\n",
    "The baroque pandasism that follows handles that."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "9429d395-668d-40ea-9406-211eb5a7af35",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 221 ms, sys: 4.04 ms, total: 225 ms\n",
      "Wall time: 225 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>timestamp</th>\n",
       "      <th>index_vec</th>\n",
       "      <th>pid_hash</th>\n",
       "      <th>hostname</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2023-11-05 00:00:00-05:00</td>\n",
       "      <td>[(28614, 1), (29460, 1), (26509, 1), (29469, 1...</td>\n",
       "      <td>[56CDA2FBD2E880CEE9520F06743C04EA, AD02784D740...</td>\n",
       "      <td>ACME-DC1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2023-11-05 00:00:00-05:00</td>\n",
       "      <td>[(30162, 1), (28614, 1), (29476, 1), (29480, 1...</td>\n",
       "      <td>[F1F1B598005B90EB45FBF77E74DDFCB8, 72DFD720300...</td>\n",
       "      <td>ACME-HH-AKA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2023-11-05 00:00:00-05:00</td>\n",
       "      <td>[(30162, 1), (29476, 1), (29446, 1), (28614, 1...</td>\n",
       "      <td>[696BA71E6CF6879C2C91E204C3C137FB, 32C0E6E988D...</td>\n",
       "      <td>ACME-HH-AXF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2023-11-05 00:00:00-05:00</td>\n",
       "      <td>[(30162, 1), (29476, 1), (29446, 1), (6953, 1)...</td>\n",
       "      <td>[62D054718CCEB3E1980225F8870D7A24, 09FC9FB7B86...</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2023-11-05 00:00:00-05:00</td>\n",
       "      <td>[(29469, 2), (29413, 2), (30162, 1), (29476, 1...</td>\n",
       "      <td>[6CFC82468479CA2C930C395DEB7E3900, A2779663E7F...</td>\n",
       "      <td>ACME-HH-CCF</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2544</th>\n",
       "      <td>2023-11-20 15:00:00-05:00</td>\n",
       "      <td>[(26509, 6), (29460, 6), (28098, 3), (29446, 3...</td>\n",
       "      <td>[9B184238C4B3508C44A0889548CE38EA, 8C4F00E4E8B...</td>\n",
       "      <td>ACME-HH-IQS</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2545</th>\n",
       "      <td>2023-11-20 18:00:00-05:00</td>\n",
       "      <td>[(3110, 1), (6222, 1), (29460, 1), (26509, 1),...</td>\n",
       "      <td>[8DDE6F47DB90619F86BD14A16F98CEE2, 9BD3697DCEA...</td>\n",
       "      <td>ACME-DC1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2546</th>\n",
       "      <td>2023-11-20 18:00:00-05:00</td>\n",
       "      <td>[(28614, 2), (28900, 1), (28098, 1), (29451, 1...</td>\n",
       "      <td>[5BB26EE4E63E3CE44E7FFAB5A0E6E321, AE736B4B16B...</td>\n",
       "      <td>ACME-HH-AZH</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2547</th>\n",
       "      <td>2023-11-20 18:00:00-05:00</td>\n",
       "      <td>[(28614, 2), (28900, 1), (28098, 1), (29451, 1...</td>\n",
       "      <td>[2B81182F7C23BE3D4A934313D3FCB52B, A15DA5AF705...</td>\n",
       "      <td>ACME-HH-IKA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2548</th>\n",
       "      <td>2023-11-20 18:00:00-05:00</td>\n",
       "      <td>[(29460, 2), (26509, 2), (28900, 1), (28098, 1...</td>\n",
       "      <td>[12024BBD1C6BF333A8D68B315C2AA06F, 287F57DF88D...</td>\n",
       "      <td>ACME-HH-IQS</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2549 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                     timestamp  \\\n",
       "0    2023-11-05 00:00:00-05:00   \n",
       "1    2023-11-05 00:00:00-05:00   \n",
       "2    2023-11-05 00:00:00-05:00   \n",
       "3    2023-11-05 00:00:00-05:00   \n",
       "4    2023-11-05 00:00:00-05:00   \n",
       "...                        ...   \n",
       "2544 2023-11-20 15:00:00-05:00   \n",
       "2545 2023-11-20 18:00:00-05:00   \n",
       "2546 2023-11-20 18:00:00-05:00   \n",
       "2547 2023-11-20 18:00:00-05:00   \n",
       "2548 2023-11-20 18:00:00-05:00   \n",
       "\n",
       "                                              index_vec  \\\n",
       "0     [(28614, 1), (29460, 1), (26509, 1), (29469, 1...   \n",
       "1     [(30162, 1), (28614, 1), (29476, 1), (29480, 1...   \n",
       "2     [(30162, 1), (29476, 1), (29446, 1), (28614, 1...   \n",
       "3     [(30162, 1), (29476, 1), (29446, 1), (6953, 1)...   \n",
       "4     [(29469, 2), (29413, 2), (30162, 1), (29476, 1...   \n",
       "...                                                 ...   \n",
       "2544  [(26509, 6), (29460, 6), (28098, 3), (29446, 3...   \n",
       "2545  [(3110, 1), (6222, 1), (29460, 1), (26509, 1),...   \n",
       "2546  [(28614, 2), (28900, 1), (28098, 1), (29451, 1...   \n",
       "2547  [(28614, 2), (28900, 1), (28098, 1), (29451, 1...   \n",
       "2548  [(29460, 2), (26509, 2), (28900, 1), (28098, 1...   \n",
       "\n",
       "                                               pid_hash     hostname  \n",
       "0     [56CDA2FBD2E880CEE9520F06743C04EA, AD02784D740...     ACME-DC1  \n",
       "1     [F1F1B598005B90EB45FBF77E74DDFCB8, 72DFD720300...  ACME-HH-AKA  \n",
       "2     [696BA71E6CF6879C2C91E204C3C137FB, 32C0E6E988D...  ACME-HH-AXF  \n",
       "3     [62D054718CCEB3E1980225F8870D7A24, 09FC9FB7B86...  ACME-HH-AZH  \n",
       "4     [6CFC82468479CA2C930C395DEB7E3900, A2779663E7F...  ACME-HH-CCF  \n",
       "...                                                 ...          ...  \n",
       "2544  [9B184238C4B3508C44A0889548CE38EA, 8C4F00E4E8B...  ACME-HH-IQS  \n",
       "2545  [8DDE6F47DB90619F86BD14A16F98CEE2, 9BD3697DCEA...     ACME-DC1  \n",
       "2546  [5BB26EE4E63E3CE44E7FFAB5A0E6E321, AE736B4B16B...  ACME-HH-AZH  \n",
       "2547  [2B81182F7C23BE3D4A934313D3FCB52B, A15DA5AF705...  ACME-HH-IKA  \n",
       "2548  [12024BBD1C6BF333A8D68B315C2AA06F, 287F57DF88D...  ACME-HH-IQS  \n",
       "\n",
       "[2549 rows x 4 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "processes_x_host_time = pd.concat([\n",
    "    (\n",
    "        group\n",
    "        .resample(pd.Timedelta(hours=3), on=\"timestamp\", origin=pd.Timestamp(\"2023-11-05\", tz=\"-0500\"))\n",
    "        .agg({\n",
    "            \"index_vec\": lambda x: list(pd.Series(x).value_counts().items()) or np.nan,\n",
    "            \"pid_hash\": lambda x: list(x) or np.nan\n",
    "        })\n",
    "        .assign(hostname=hostname)\n",
    "        .reset_index()\n",
    "    )\n",
    "    for hostname, group in processes_indexed.groupby(\"hostname\")\n",
    "]).dropna().sort_values([\"timestamp\", \"hostname\"]).reset_index(drop=True)\n",
    "processes_x_host_time"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2964710e-fe91-4609-8218-8092cc4cdf62",
   "metadata": {},
   "source": [
    "Ok, so we have the command line vectors, what we are missing are the weighted distributions.\n",
    "Each row of `processes_x_host_time` above represents a (host, time interval) couple &mdash; let's call these _host-times_.\n",
    "The `index_vec` column contains a list of (command line index, count) for each of these host-times.\n",
    "Since host-times are expressed over the vocabulary of (vector-represented) command lines,\n",
    "these are essentially the elements we need to build the count matrix.\n",
    "If we [explode](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.explode.html) the `index_vec` column,\n",
    "each row of the resulting data frame contains a triple of the form\n",
    "\n",
    "```\n",
    "(row identifier, column identifier, count)\n",
    "```\n",
    "\n",
    "A list of such triples suffices to build up the count matrix in a [sparse COO format](https://matteding.github.io/2019/04/25/sparse-matrices/#coordinate-matrix).\n",
    "We have the `vectorizers.EdgeListVectorizer` to render this.\n",
    "Notice that we are forcing the column space to match the full set of command line indices.\n",
    "This is because otherwise, `EdgeListVectorizer` realizes that some command line indices are not used between minimum and maximum observed indices, and discards the corresponding columns.\n",
    "Forcing the existence of these empty columns makes it easier to match counts back to our vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "fa06e53f-cf37-44f3-a53d-74c93c836983",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 14.6 s, sys: 46.9 ms, total: 14.6 s\n",
      "Wall time: 14.6 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<2549x30994 sparse matrix of type '<class 'numpy.float64'>'\n",
       "\twith 74600 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "vz_tc = vz.EdgeListVectorizer(column_label_dictionary={i: i for i in range(len(cmdlines_vec))}).fit([\n",
    "    ((t.timestamp, t.hostname), *t.index_vec)\n",
    "    for t in processes_x_host_time.explode(\"index_vec\").dropna().itertuples(index=False)\n",
    "])\n",
    "pxht_tc = vz_tc._train_matrix.tocsr()\n",
    "pxht_tc"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "075fb34f-10e2-4eb3-81d8-2c242304cc96",
   "metadata": {},
   "source": [
    "Modulate counts by information contribution."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "eb109c75-5e34-4ce6-86cd-d422d1eda440",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 337 ms, sys: 12.2 ms, total: 350 ms\n",
      "Wall time: 351 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<2549x30994 sparse matrix of type '<class 'numpy.float64'>'\n",
       "\twith 74600 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "pxht_iwt = vzt.InformationWeightTransformer().fit_transform(pxht_tc)\n",
    "pxht_iwt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "686eb1fa-bb1d-4ac1-a86a-c3eb53ca8ec6",
   "metadata": {},
   "source": [
    "Now we have what we need to vectorize the host-times."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2919a904-f233-4bbe-a9ae-89e42b0b9f2d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 20s, sys: 8.25 s, total: 1min 29s\n",
      "Wall time: 17.8 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 4.65558449e+00, -5.58356032e+00, -1.12563740e+00, ...,\n",
       "        -7.68703067e-03, -4.45136702e-01, -1.57195824e-01],\n",
       "       [ 2.32172237e+00, -1.87163266e+00, -6.17892512e-02, ...,\n",
       "         1.32848456e-01, -6.85284632e-02,  6.81173192e-01],\n",
       "       [ 4.44249311e+00, -2.32975679e+00, -4.78578250e-01, ...,\n",
       "        -7.46442906e-02,  5.27855187e-01, -4.55018328e-01],\n",
       "       ...,\n",
       "       [ 5.57285201e+00, -4.92374956e+00,  4.59681744e-02, ...,\n",
       "         1.07394110e-01,  1.15181618e-01, -4.20507171e-01],\n",
       "       [ 6.96246260e+00, -3.61103338e+00, -6.53831942e-01, ...,\n",
       "         1.41542502e-01,  5.01234611e-01, -6.76426525e-01],\n",
       "       [ 5.60750526e+00, -4.76178381e+00, -2.83156726e-01, ...,\n",
       "        -3.18677832e-03,  2.22560767e-01, -2.43053562e-01]])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "pxht_vec = vz.WassersteinVectorizer().fit_transform(pxht_iwt, vectors=cmdlines_vec)\n",
    "pxht_vec"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "43b00e62-a975-4390-ac13-d1b135c13433",
   "metadata": {},
   "source": [
    "Compress these vectors into a data map."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b419fd50-ba88-41d9-93a1-827717320a1c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "UMAP(angular_rp_forest=True, metric='cosine', verbose=True)\n",
      "Tue Jul  9 21:53:14 2024 Construct fuzzy simplicial set\n",
      "Tue Jul  9 21:53:16 2024 Finding Nearest Neighbors\n",
      "Tue Jul  9 21:53:17 2024 Finished Nearest Neighbor Search\n",
      "Tue Jul  9 21:53:18 2024 Construct embedding\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5a291040bdc8476a9abc98e4040e5244",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Epochs completed:   0%|            0/500 [00:00]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\tcompleted  0  /  500 epochs\n",
      "\tcompleted  50  /  500 epochs\n",
      "\tcompleted  100  /  500 epochs\n",
      "\tcompleted  150  /  500 epochs\n",
      "\tcompleted  200  /  500 epochs\n",
      "\tcompleted  250  /  500 epochs\n",
      "\tcompleted  300  /  500 epochs\n",
      "\tcompleted  350  /  500 epochs\n",
      "\tcompleted  400  /  500 epochs\n",
      "\tcompleted  450  /  500 epochs\n",
      "Tue Jul  9 21:53:22 2024 Finished embedding\n",
      "CPU times: user 28.2 s, sys: 2.64 s, total: 30.9 s\n",
      "Wall time: 8.43 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "pxht_dmap = umap.UMAP(metric=\"cosine\", verbose=True).fit_transform(pxht_vec)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "0380de95-8972-4967-8e89-c4252381a4b8",
   "metadata": {},
   "source": [
    "As I said [above](#interpretation),\n",
    "points in the data map get close together when the corresponding host-times are characterized by similar ~processes~ command lines.\n",
    "So it would be nice to annotate the clusters of the data map so we may understand the commands or programs that make them up.\n",
    "I think of it as _descriptions of workloads._\n",
    "Unfortunately, there is not much room in a data map to describe the multiple programs that make up a workload.\n",
    "Let's try compressing that to the _3 most important processes_, characterized by their name."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "17b7d607-a8e5-486f-bfe0-71eac2807872",
   "metadata": {},
   "source": [
    "How can we measure the importance of a program to a workload?\n",
    "The information-reweighted count matrix can answer this question.\n",
    "The information weight transformer replaces raw counts with _numbers of bits of information_ contributed by each component that characterizes a bag.\n",
    "One takes the host-times that make up a cluster, sums their corresponding rows of the IWT matrix.\n",
    "Then take each such sum in turn, in decreasing order:\n",
    "any large number of bits corresponds to a command line that is significant for characterizing this host-time.\n",
    "Then find all other non-zero-bits process instances of the same name as this one high-information command line\n",
    "(within the same cluster),\n",
    "combine them together.\n",
    "If there are enough bits between them all, then this program truly makes this cluster accrete together, and we use it to annotate the cluster.\n",
    "We take as much as 3 such program names to form the label."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d3128a07-194f-42e2-b537-fedc863d64cd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "162"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clusters = HDBSCAN(cluster_selection_method=\"leaf\").fit_predict(pxht_dmap)\n",
    "num_clusters = np.unique(clusters).shape[0] - 1\n",
    "num_clusters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "96b71471-aa24-4b3b-a318-3455291db8ac",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "14655b9e57234689a97a3ebe71f19008",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/162 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>labels_cluster</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>upfc.exe 2.9\\nsihclient.exe 2.9\\nsc.exe 2.1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>microsoftedge_x64_119.0.2151.72_119.0.2151.58....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>microsoftedge_x64_119.0.2151.44_118.0.2088.76....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>upfc.exe 2.8\\ndismhost.exe 2.7\\ncleanmgr.exe 2.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>googleupdate.exe 9.0\\nsihclient.exe 2.7\\nupfc....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157</th>\n",
       "      <td>ngen.exe 2.3\\nsearchprotocolhost.exe 2.2\\ndism...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>158</th>\n",
       "      <td>dismhost.exe 2.1\\nmsedge.exe 2.0\\ncompattelrun...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>159</th>\n",
       "      <td>dismhost.exe 2.1\\nmsedge.exe 2.1\\ndevicecensus...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>160</th>\n",
       "      <td>msedge.exe 2.2\\nbackgroundtaskhost.exe 1.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161</th>\n",
       "      <td>backgroundtaskhost.exe 4.3\\ncalc.exe 3.1\\nwin3...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>162 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                        labels_cluster\n",
       "0          upfc.exe 2.9\\nsihclient.exe 2.9\\nsc.exe 2.1\n",
       "1    microsoftedge_x64_119.0.2151.72_119.0.2151.58....\n",
       "2    microsoftedge_x64_119.0.2151.44_118.0.2088.76....\n",
       "3     upfc.exe 2.8\\ndismhost.exe 2.7\\ncleanmgr.exe 2.6\n",
       "4    googleupdate.exe 9.0\\nsihclient.exe 2.7\\nupfc....\n",
       "..                                                 ...\n",
       "157  ngen.exe 2.3\\nsearchprotocolhost.exe 2.2\\ndism...\n",
       "158  dismhost.exe 2.1\\nmsedge.exe 2.0\\ncompattelrun...\n",
       "159  dismhost.exe 2.1\\nmsedge.exe 2.1\\ndevicecensus...\n",
       "160         msedge.exe 2.2\\nbackgroundtaskhost.exe 1.2\n",
       "161  backgroundtaskhost.exe 4.3\\ncalc.exe 3.1\\nwin3...\n",
       "\n",
       "[162 rows x 1 columns]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels_cluster = {}\n",
    "last_bits = 0\n",
    "for cluster in trange(num_clusters):\n",
    "    iwt_cluster = np.array(pxht_iwt[clusters == cluster, :].todense())\n",
    "    bits_cluster = np.sum(iwt_cluster, axis=0)\n",
    "    processes_cluster = (\n",
    "        processes_indexed\n",
    "        .merge(\n",
    "            processes_x_host_time[\"pid_hash\"].iloc[clusters == cluster].explode(),\n",
    "            on=\"pid_hash\",\n",
    "            how=\"inner\"\n",
    "        )\n",
    "        .merge(\n",
    "            pd.Series(bits_cluster, name=\"bits\"),\n",
    "            how=\"inner\",\n",
    "            left_on=\"index_vec\",\n",
    "            right_index=True\n",
    "        )\n",
    "    )\n",
    "\n",
    "    i_important = np.argsort(-bits_cluster)\n",
    "    names = {}\n",
    "    for i in i_important:\n",
    "        if bits_cluster[i] < 1e-6:\n",
    "            break\n",
    "        processes_cmdline = processes_cluster.loc[processes_cluster[\"index_vec\"] == i]\n",
    "        assert not processes_cmdline.empty\n",
    "        name = processes_cmdline.iloc[0][\"name\"]\n",
    "        if name not in names:\n",
    "            processes_same_name = processes_cluster.loc[processes_cluster[\"name\"] == name]\n",
    "            bits = processes_same_name[\"bits\"].mean()\n",
    "            if bits >= 1.:\n",
    "                names[name] = bits\n",
    "    if names:\n",
    "        names_sorted = sorted(list(names.items()), key=lambda p: p[1], reverse=True)\n",
    "        _, bits_max = names_sorted[0]\n",
    "        names_filtered = [(name, bits) for name, bits in names_sorted if bits >= bits_max / 10.]\n",
    "        labels_cluster[cluster] = \"\\n\".join(\n",
    "            f\"{name} {bits:.1f}\"\n",
    "            for name, bits in names_filtered[:3]\n",
    "        )\n",
    "    else:\n",
    "        labels_cluster[cluster] = \"Unlabelled\"\n",
    "pd.DataFrame(labels_cluster.values(), columns=[\"labels_cluster\"], index=labels_cluster.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "669cd7eb-1be3-4b49-bd41-7a2c675a0aa8",
   "metadata": {},
   "source": [
    "Propagate these cluster labels to the points."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "5f9f635b-c423-4d4f-a8fc-6952e259d629",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0                         svchost.exe 3.4\\nmsedge.exe 2.7\n",
       "1                                        dsregcmd.exe 6.5\n",
       "2       speechmodeldownload.exe 3.5\\nmusnotification.e...\n",
       "3                              backgroundtaskhost.exe 1.2\n",
       "4       speechmodeldownload.exe 1.4\\nmusnotification.e...\n",
       "                              ...                        \n",
       "2544    mpcmdrun.exe 2.8\\nbackgroundtaskhost.exe 2.6\\n...\n",
       "2545    svchost.exe 3.2\\nmsedge.exe 2.7\\nmergehelper.e...\n",
       "2546                                           sc.exe 3.0\n",
       "2547                                           Unlabelled\n",
       "2548    sc.exe 1.8\\nbackgroundtaskhost.exe 1.3\\nrundll...\n",
       "Length: 2549, dtype: object"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels = [labels_cluster.get(c, \"\") or \"Unlabelled\" for c in clusters]\n",
    "pd.Series(labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7fbe5ddf-6be3-4c3e-8531-4e912a8ee932",
   "metadata": {},
   "source": [
    "We get to identify host-times quite simply: with hostname and, um, time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "c91d11f4-f5d4-4f21-bedc-5a2cbf98abab",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0          ACME-DC1, Nov 05, 00:00\n",
       "1       ACME-HH-AKA, Nov 05, 00:00\n",
       "2       ACME-HH-AXF, Nov 05, 00:00\n",
       "3       ACME-HH-AZH, Nov 05, 00:00\n",
       "4       ACME-HH-CCF, Nov 05, 00:00\n",
       "                   ...            \n",
       "2544    ACME-HH-IQS, Nov 20, 15:00\n",
       "2545       ACME-DC1, Nov 20, 18:00\n",
       "2546    ACME-HH-AZH, Nov 20, 18:00\n",
       "2547    ACME-HH-IKA, Nov 20, 18:00\n",
       "2548    ACME-HH-IQS, Nov 20, 18:00\n",
       "Length: 2549, dtype: object"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hover_text = processes_x_host_time[\"hostname\"] + \", \" + processes_x_host_time.timestamp.dt.strftime(\"%b %d, %H:%M\")\n",
    "hover_text"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99f54eba-d0f9-49a8-94a2-a969e7f0a11e",
   "metadata": {},
   "source": [
    "So far, we have let labeling determine the color of the points.\n",
    "Since time plays an important role here, let's use point color to track it,\n",
    "using a continuous color map."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "8e8b67bc-ea9e-4fb5-8dd0-4310d2b7c51a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/html": [
       "<div style=\"vertical-align: middle;\"><strong>cool</strong> </div><div class=\"cmap\"><img alt=\"cool colormap\" title=\"cool\" style=\"border: 1px solid #555;\" src=\"data:image/png;base64,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\"></div><div style=\"vertical-align: middle; max-width: 514px; display: flex; justify-content: space-between;\"><div style=\"float: left;\"><div title=\"#00ffffff\" style=\"display: inline-block; width: 1em; height: 1em; margin: 0; vertical-align: middle; border: 1px solid #555; background-color: #00ffffff;\"></div> under</div><div style=\"margin: 0 auto; display: inline-block;\">bad <div title=\"#00000000\" style=\"display: inline-block; width: 1em; height: 1em; margin: 0; vertical-align: middle; border: 1px solid #555; background-color: #00000000;\"></div></div><div style=\"float: right;\">over <div title=\"#ff00ffff\" style=\"display: inline-block; width: 1em; height: 1em; margin: 0; vertical-align: middle; border: 1px solid #555; background-color: #ff00ffff;\"></div></div>"
      ],
      "text/plain": [
       "<matplotlib.colors.LinearSegmentedColormap at 0x17f2698b0>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cmap = colormaps[\"cool\"]\n",
    "cmap"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba397c43-aa8c-42a4-888a-b75c5930f370",
   "metadata": {},
   "source": [
    "So cool points correspond to early host-times, and hot ones, to late host-times."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "e1c0f37b-b5b4-49f1-9356-e5329ea5c015",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "            <iframe\n",
       "                width=100%\n",
       "                height=800\n",
       "                frameborder=\"0\"\n",
       "                srcdoc=\"\n",
       "&lt;!DOCTYPE html&gt;\n",
       "&lt;html&gt;\n",
       "  &lt;head&gt;\n",
       "    &lt;meta http-equiv=&quot;content-type&quot; content=&quot;text/html; charset=UTF-8&quot; /&gt;\n",
       "    &lt;title&gt;Hosts over time&lt;/title&gt;\n",
       "    \n",
       "    &lt;link rel=&quot;preconnect&quot; href=&quot;https://fonts.googleapis.com&quot;&gt;\n",
       "    &lt;link rel=&quot;preconnect&quot; href=&quot;https://fonts.gstatic.com&quot; crossorigin&gt;\n",
       "    &lt;link href=&quot;https://fonts.googleapis.com/css2?family=Roboto&amp;display=swap&quot; rel=&quot;stylesheet&quot;&gt;\n",
       "    \n",
       "    \n",
       "       \n",
       "    &lt;link rel=&quot;stylesheet&quot; href=&quot;https://maxcdn.bootstrapcdn.com/bootstrap/3.2.0/css/bootstrap-theme.min.css&quot; /&gt;\n",
       "    &lt;link rel=&quot;stylesheet&quot; href=&quot;https://maxcdn.bootstrapcdn.com/font-awesome/4.6.3/css/font-awesome.min.css&quot; /&gt;\n",
       "    &lt;script src=&quot;https://unpkg.com/deck.gl@latest/dist.min.js&quot;&gt;&lt;/script&gt;\n",
       "    \n",
       "    &lt;script src=&quot;https://unpkg.com/fflate@0.8.0&quot;&gt;&lt;/script&gt;\n",
       "    \n",
       "    &lt;style&gt;\n",
       "        body {\n",
       "          margin: 0;\n",
       "          padding: 0;\n",
       "          overflow: hidden;\n",
       "          background: #000000;\n",
       "        }\n",
       "\n",
       "        #deck-container {\n",
       "          width: 100vw;\n",
       "          height: 100vh;\n",
       "        }\n",
       "\n",
       "        #deck-container canvas {\n",
       "          z-index: 1;\n",
       "          background: #000000;\n",
       "        }\n",
       "\n",
       "        .deck-tooltip {\n",
       "            \n",
       "            font-size: 0.8em;\n",
       "            font-family: Roboto;\n",
       "            font-weight: 300;\n",
       "            color: #ffffff !important;\n",
       "            background-color: #000000aa !important;\n",
       "            border-radius: 12px;\n",
       "            box-shadow: 2px 3px 10px #00000044;\n",
       "            max-width: 25%;\n",
       "        }\n",
       "        \n",
       "        #loading {\n",
       "            width: 100%;\n",
       "            height: 100%;\n",
       "            top: 0px;\n",
       "            left: 0px;\n",
       "            position: absolute;\n",
       "            display: block; \n",
       "            z-index: 99\n",
       "        }\n",
       "\n",
       "        #loading-image {\n",
       "            position: absolute;\n",
       "            top: 45%;\n",
       "            left: 47.5%;\n",
       "            z-index: 100\n",
       "        }\n",
       "        \n",
       "        #title-container {\n",
       "            position: absolute;\n",
       "            top: 0;\n",
       "            left: 0;\n",
       "            margin: 16px;\n",
       "            padding: 12px;\n",
       "            border-radius: 16px;\n",
       "            line-height: 0.95;\n",
       "            z-index: 2;\n",
       "            font-family: Roboto;\n",
       "            color: #ffffff;\n",
       "            background: #000000aa;\n",
       "            box-shadow: 2px 3px 10px #00000044;\n",
       "        }\n",
       "        \n",
       "        \n",
       "        #search-container{\n",
       "            position: absolute;\n",
       "            left: -16px;\n",
       "            margin: 16px;\n",
       "            padding: 12px;\n",
       "            border-radius: 16px;\n",
       "            z-index: 2;\n",
       "            font-family: Roboto;\n",
       "            color: #ffffff;\n",
       "            background: #000000aa;\n",
       "            width: fit-content;\n",
       "            box-shadow: 2px 3px 10px #00000044;\n",
       "        }\n",
       "        input {\n",
       "            margin: 2px;\n",
       "            padding: 4px;\n",
       "            border-radius: 8px;\n",
       "            color: #ffffff;\n",
       "            background: #000000dd;\n",
       "            border: 1px solid 222222ff;\n",
       "            transition: 0.5s;\n",
       "            outline: none;\n",
       "        }\n",
       "        input:focus {\n",
       "          border: 2px solid #555;\n",
       "        }\n",
       "        \n",
       "        \n",
       "    &lt;/style&gt;\n",
       "  &lt;/head&gt;\n",
       "  &lt;body&gt;\n",
       "    &lt;div id=&quot;loading&quot;&gt;\n",
       "        &lt;img id=&quot;loading-image&quot; src=&quot;https://i.gifer.com/ZKZg.gif&quot; alt=&quot;Loading...&quot; width=&quot;5%&quot;/&gt;\n",
       "    &lt;/div&gt;\n",
       "    \n",
       "    &lt;div id=&quot;title-container&quot;&gt;\n",
       "        &lt;span style=&quot;font-family:Roboto;font-size:36pt;color:#ffffff&quot;&gt;\n",
       "            Hosts over time\n",
       "        &lt;/span&gt;&lt;br/&gt;\n",
       "        &lt;span style=&quot;font-family:Roboto;font-size:18pt;color:#777777&quot;&gt;\n",
       "            as distributions over a cloud of command line vectors\n",
       "        &lt;/span&gt;\n",
       "        \n",
       "        &lt;div id=&quot;search-container&quot;&gt;\n",
       "            &lt;input autocomplete=&quot;off&quot; type=&quot;search&quot; id=&quot;search&quot; placeholder=&quot;🔍&quot;&gt;\n",
       "        &lt;/div&gt;\n",
       "        \n",
       "    &lt;/div&gt;\n",
       "    \n",
       "    \n",
       "    &lt;div id=&quot;deck-container&quot;&gt;\n",
       "    &lt;/div&gt;\n",
       "    \n",
       "  &lt;/body&gt;\n",
       "  &lt;script type=&quot;module&quot;&gt;\n",
       "    import { ArrowLoader } from &#x27;https://cdn.jsdelivr.net/npm/@loaders.gl/arrow@4.1.0-alpha.10/+esm&#x27;\n",
       "    import { JSONLoader } from &#x27;https://cdn.jsdelivr.net/npm/@loaders.gl/json@4.0.5/+esm&#x27;\n",
       "    \n",
       "\n",
       "    \n",
       "    const pointDataBase64 = &quot;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&quot;;\n",
       "    const pointDataBuffer = fflate.strToU8(atob(pointDataBase64), true);\n",
       "    const pointData = await loaders.parse(pointDataBuffer, ArrowLoader);\n",
       "    const hoverDataBase64 = &quot;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&quot;;\n",
       "    const hoverDataBuffer = fflate.strToU8(atob(hoverDataBase64), true);\n",
       "    const unzippedHoverData = fflate.gunzipSync(hoverDataBuffer);\n",
       "    const hoverData = await loaders.parse(unzippedHoverData, ArrowLoader);\n",
       "    const labelDataBase64 = &quot;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&quot;;\n",
       "    const labelDataBuffer = fflate.strToU8(atob(labelDataBase64), true);\n",
       "    const unzippedLabelData = fflate.gunzipSync(labelDataBuffer);    \n",
       "    const labelData = await loaders.parse(unzippedLabelData, JSONLoader);\n",
       "    \n",
       "    \n",
       "    const DATA = {src: pointData.data, length: pointData.data.x.length}\n",
       "\n",
       "    const container = document.getElementById(&#x27;deck-container&#x27;);\n",
       "    const pointLayer = new deck.ScatterplotLayer({\n",
       "        id: &#x27;dataPointLayer&#x27;,\n",
       "        data: DATA,\n",
       "        getPosition: (object, {index, data}) =&gt; {\n",
       "            return [data.src.x[index], data.src.y[index]];\n",
       "        },\n",
       "        \n",
       "        getRadius: 0.1,\n",
       "        \n",
       "        getFillColor: (object, {index, data}) =&gt; {\n",
       "            return [\n",
       "                data.src.r[index], \n",
       "                data.src.g[index], \n",
       "                data.src.b[index],\n",
       "                180\n",
       "            ]\n",
       "        },\n",
       "        getLineColor: (object, {index, data}) =&gt; {\n",
       "            return [\n",
       "                data.src.r[index], \n",
       "                data.src.g[index], \n",
       "                data.src.b[index],\n",
       "                32\n",
       "            ]\n",
       "        },       \n",
       "        getLineColor: [5, 5, 5, 128],\n",
       "        getLineWidth: 0.001,\n",
       "        highlightColor: [170, 0, 0, 187], \n",
       "        lineWidthMaxPixels: 8,\n",
       "        lineWidthMinPixels: 0.1,\n",
       "        radiusMaxPixels: 24, \n",
       "        radiusMinPixels: 0.01,\n",
       "        radiusUnits: &quot;common&quot;, \n",
       "        lineWidthUnits: &quot;common&quot;, \n",
       "        autoHighlight: true,\n",
       "        pickable: true, \n",
       "        stroked: true\n",
       "    });\n",
       "    const labelLayer = new deck.TextLayer({\n",
       "        id: &quot;textLabelLayer&quot;,\n",
       "        data: labelData,\n",
       "        pickable: false,\n",
       "        getPosition: d =&gt; [d.x, d.y],\n",
       "        getText: d =&gt; d.label,\n",
       "        getColor: [255, 255, 255, 255],\n",
       "        getSize: d =&gt; d.size,\n",
       "        sizeScale: 1,\n",
       "        sizeMinPixels: 18,\n",
       "        sizeMaxPixels: 36,\n",
       "        outlineWidth: 8,\n",
       "        outlineColor: [17, 17, 17, 221],\n",
       "        getBackgroundColor: [0, 0, 0, 64],\n",
       "        getBackgroundPadding: [15, 15, 15, 15],\n",
       "        background: true,\n",
       "        characterSet: &quot;auto&quot;,\n",
       "        fontFamily: &quot;Roboto&quot;,\n",
       "        fontWeight: 900,\n",
       "        lineHeight: 0.95,\n",
       "        fontSettings: {&quot;sdf&quot;: true},\n",
       "        getTextAnchor: &quot;middle&quot;,\n",
       "        getAlignmentBaseline: &quot;center&quot;,\n",
       "        lineHeight: 0.95,\n",
       "        elevation: 100,\n",
       "        // CollideExtension options\n",
       "        collisionEnabled: true,\n",
       "        getCollisionPriority: d =&gt; d.size,\n",
       "        collisionTestProps: {\n",
       "          sizeScale: 3,\n",
       "          sizeMaxPixels: 36 * 2,\n",
       "          sizeMinPixels: 18 * 2\n",
       "        },\n",
       "        extensions: [new deck.CollisionFilterExtension()],\n",
       "    });\n",
       "    \n",
       "\n",
       "    const deckgl = new deck.DeckGL({\n",
       "      container: container,\n",
       "      initialViewState: {\n",
       "        latitude: 3.9935603,\n",
       "        longitude: 4.0839114,\n",
       "        zoom: 3.889756586091381\n",
       "      },\n",
       "      controller: true,\n",
       "      \n",
       "      layers: [pointLayer, labelLayer],\n",
       "      \n",
       "      \n",
       "      getTooltip: ({index}) =&gt; hoverData.data.hover_text[index]\n",
       "    });\n",
       "    \n",
       "    document.getElementById(&quot;loading&quot;).style.display = &quot;none&quot;;\n",
       "        \n",
       "    \n",
       "        function selectPoints(item, conditional) {\n",
       "        var layerId;\n",
       "        if (item) {\n",
       "            for (var i = 0; i &lt; DATA.length; i++) {\n",
       "                if (conditional(i)) {\n",
       "                    DATA.src.selected[i] = 1;\n",
       "                } else {\n",
       "                    DATA.src.selected[i] = 0;\n",
       "                }\n",
       "            }\n",
       "            layerId = &#x27;selectedPointLayer&#x27; + item;\n",
       "        } else {\n",
       "            for (var i = 0; i &lt; DATA.length; i++) {\n",
       "                DATA.src.selected[i] = 1;\n",
       "            }\n",
       "            layerId = &#x27;dataPointLayer&#x27;;\n",
       "        }\n",
       "        const selectedPointLayer = pointLayer.clone(\n",
       "            {\n",
       "                id: layerId,\n",
       "                data: DATA,\n",
       "                getFilterValue: (object, {index, data}) =&gt; data.src.selected[index],\n",
       "                filterRange: [1, 2],\n",
       "                extensions: [new deck.DataFilterExtension({filterSize: 1})]\n",
       "            }\n",
       "        );\n",
       "        deckgl.setProps(\n",
       "            {layers: \n",
       "                [selectedPointLayer].concat(deckgl.props.layers.slice(1,))\n",
       "            }\n",
       "        );\n",
       "    }\n",
       "    \n",
       "    const search = document.getElementById(&quot;search&quot;);\n",
       "    search.addEventListener(&quot;input&quot;, (event) =&gt; {\n",
       "            const search_term = event.target.value.toLowerCase();\n",
       "            selectPoints(search_term, (i) =&gt; hoverData.data.hover_text[i].toLowerCase().includes(search_term));\n",
       "        }\n",
       "    );\n",
       "    \n",
       "    \n",
       "    &lt;/script&gt;\n",
       "&lt;/html&gt;\"\n",
       "            ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<InteractiveFigure width=100% height=800>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ts_min, ts_max = processes_x_host_time[\"timestamp\"].agg([\"min\", \"max\"])\n",
    "delta = ts_max - ts_min\n",
    "dmp.create_interactive_plot(\n",
    "    pxht_dmap,\n",
    "    labels,\n",
    "    hover_text=hover_text,\n",
    "    title=\"Hosts over time\",\n",
    "    sub_title=\"as distributions over a cloud of command line vectors\",\n",
    "    darkmode=True,\n",
    "    color_label_text=False,\n",
    "    marker_color_array=[cmap((ts - ts_min) / delta) for ts in processes_x_host_time[\"timestamp\"]],\n",
    "    enable_search=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "ebbdd923-5198-4fae-8d58-c50f082b8add",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['ACME-DC1', 'ACME-HH-AKA', 'ACME-HH-AXF', 'ACME-HH-AZH',\n",
       "       'ACME-HH-CCF', 'ACME-HH-CLR', 'ACME-HH-DFN', 'ACME-HH-EAY',\n",
       "       'ACME-HH-HGC', 'ACME-HH-IQS', 'ACME-HH-MBY', 'ACME-HH-OLY',\n",
       "       'ACME-HH-TTW', 'ACME-HH-UWI', 'ACME-HH-YIM', 'ACME-WS-AZU',\n",
       "       'ACME-WS-NUQ', 'ACME-WS-PLU', 'ACME-WS-PRV', 'ACME-WS-UVF',\n",
       "       'ACME-HH-ATV', 'ACME-HH-IKA', 'ACME-HH-KKN', 'ACME-HH-TON',\n",
       "       'ACME-HH-BVW'], dtype=object)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "processes_x_host_time[\"hostname\"].unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c12aef0-9f16-4acb-b752-e14172ad0b2c",
   "metadata": {},
   "source": [
    "At a glance, we see that no software workload is time specific:\n",
    "all regions of the space host points running the whole color range.\n",
    "This visualization makes it difficult to track how hosts cover the space;\n",
    "a quick look around hovering over points suggest that most hosts\n",
    "carry workloads all over the space, given time.\n",
    "None of these hosts manifests any specialized behaviour:\n",
    "they just zoom around the space over time.\n",
    "This is surprising at first blush, given that at least one of these hosts,\n",
    "`ACME-DC1`, is the Domain Controller of the network &mdash; a rather specific role.\n",
    "The generic shape of its behaviour likely stems from the network being small and barely used compared to a regular enterprise network.\n",
    "As such, the DC services a handful of authentication queries per day,\n",
    "making it look just like any other workstation on the network."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.12.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}