SAMKNN_BoT_Conference.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "31f8ac4f-8f65-4090-9fe6-d17be790b535",
   "metadata": {},
   "outputs": [],
   "source": [
    "import copy\n",
    "import pickle\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from tqdm import tqdm \n",
    "\n",
    "from sklearn.utils import shuffle\n",
    "from sklearn.metrics import precision_score, f1_score, accuracy_score, recall_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from river import metrics, evaluate, stream, compose, drift, preprocessing \n",
    "\n",
    "from collections import defaultdict, Counter\n",
    "\n",
    "from AdaptiveSAM import AdaptiveSAMKNNClassifier\n",
    "from SAMKNN import SAMKNNClassifier\n",
    "from AdaptiveSAMPlot import AdaptiveSAMKNNClassifierPlot\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "from matplotlib.ticker import FuncFormatter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fd02c594-5e39-4999-8325-b74b22a7328f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[50,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.41%   94.93%   97.12%      9894  \n",
      "       1      98.76%   99.86%   99.31%     40105  \n",
      "                                                  \n",
      "   Macro      99.08%   97.39%   98.21%            \n",
      "   Micro      98.88%   98.88%   98.88%            \n",
      "Weighted      98.89%   98.88%   98.87%            \n",
      "\n",
      "                 98.88% accuracy                   – 00:00:08 – 63.07 KB\n",
      "[70,703]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.48%   94.80%   97.09%     13859  \n",
      "       1      98.75%   99.88%   99.31%     56843  \n",
      "                                                  \n",
      "   Macro      99.11%   97.34%   98.20%            \n",
      "   Micro      98.88%   98.88%   98.88%            \n",
      "Weighted      98.89%   98.88%   98.87%            \n",
      "\n",
      "                 98.88% accuracy                   – 00:00:13 – 64.64 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      99.48%   94.80%   97.09%     13859  \n",
       "       1      98.75%   99.88%   99.31%     56843  \n",
       "                                                  \n",
       "   Macro      99.11%   97.34%   98.20%            \n",
       "   Micro      98.88%   98.88%   98.88%            \n",
       "Weighted      98.89%   98.88%   98.87%            \n",
       "\n",
       "                 98.88% accuracy                  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT-DDoS.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model_bot_ddos = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model_bot_ddos, report, print_every=50000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "7c317c55-407b-4b9a-8a13-bf3fadd0a06b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save\n",
    "with open('model_bot_ddos_con.pkl', 'wb') as f:\n",
    "    pickle.dump(model_bot_ddos, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7ec34be8-dd91-4c7b-88eb-3951454667af",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[50,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.62%   94.73%   97.11%      9881  \n",
      "       1      98.72%   99.91%   99.31%     40118  \n",
      "                                                  \n",
      "   Macro      99.17%   97.32%   98.21%            \n",
      "   Micro      98.89%   98.89%   98.89%            \n",
      "Weighted      98.89%   98.89%   98.88%            \n",
      "\n",
      "                 98.89% accuracy                   – 00:00:09 – 63.07 KB\n",
      "[70,692]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.58%   94.80%   97.13%     13859  \n",
      "       1      98.75%   99.90%   99.32%     56832  \n",
      "                                                  \n",
      "   Macro      99.16%   97.35%   98.23%            \n",
      "   Micro      98.90%   98.90%   98.90%            \n",
      "Weighted      98.91%   98.90%   98.89%            \n",
      "\n",
      "                 98.90% accuracy                   – 00:00:13 – 64.62 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      99.58%   94.80%   97.13%     13859  \n",
       "       1      98.75%   99.90%   99.32%     56832  \n",
       "                                                  \n",
       "   Macro      99.16%   97.35%   98.23%            \n",
       "   Micro      98.90%   98.90%   98.90%            \n",
       "Weighted      98.91%   98.90%   98.89%            \n",
       "\n",
       "                 98.90% accuracy                  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT-DoS.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model_bot_dos = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model_bot_dos, report, print_every=50000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ae0e09d2-a818-4dc5-84ba-fdb21326e1e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save\n",
    "with open('model_bot_dos_con.pkl', 'wb') as f:\n",
    "    pickle.dump(model_bot_dos, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "10fabc71-92e9-4045-a2d2-7747fd989989",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[50,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.91%   30.90%   43.40%      1437  \n",
      "       1      97.99%   99.66%   98.82%     48562  \n",
      "                                                  \n",
      "   Macro      85.45%   65.28%   71.11%            \n",
      "   Micro      97.68%   97.68%   97.68%            \n",
      "Weighted      97.27%   97.68%   97.23%            \n",
      "\n",
      "                 97.68% accuracy                   – 00:00:08 – 63.09 KB\n",
      "[100,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.24%   30.83%   43.57%      2861  \n",
      "       1      98.00%   99.68%   98.83%     97138  \n",
      "                                                  \n",
      "   Macro      86.12%   65.26%   71.20%            \n",
      "   Micro      97.71%   97.71%   97.71%            \n",
      "Weighted      97.32%   97.71%   97.25%            \n",
      "\n",
      "                 97.71% accuracy                   – 00:00:18 – 63.09 KB\n",
      "[150,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.67%   31.66%   44.47%      4292  \n",
      "       1      98.02%   99.68%   98.85%    145707  \n",
      "                                                  \n",
      "   Macro      86.35%   65.67%   71.66%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      97.35%   97.74%   97.29%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:00:27 – 64.64 KB\n",
      "[200,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.58%   31.48%   44.27%      5706  \n",
      "       1      98.02%   99.69%   98.85%    194293  \n",
      "                                                  \n",
      "   Macro      86.30%   65.58%   71.56%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      97.35%   97.74%   97.29%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:00:36 – 64.64 KB\n",
      "[250,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      75.02%   31.82%   44.69%      7096  \n",
      "       1      98.04%   99.69%   98.86%    242903  \n",
      "                                                  \n",
      "   Macro      86.53%   65.76%   71.77%            \n",
      "   Micro      97.76%   97.76%   97.76%            \n",
      "Weighted      97.39%   97.76%   97.32%            \n",
      "\n",
      "                 97.76% accuracy                   – 00:00:45 – 63.09 KB\n",
      "[300,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      75.19%   32.42%   45.31%      8599  \n",
      "       1      98.04%   99.68%   98.85%    291400  \n",
      "                                                  \n",
      "   Macro      86.61%   66.05%   72.08%            \n",
      "   Micro      97.76%   97.76%   97.76%            \n",
      "Weighted      97.38%   97.76%   97.32%            \n",
      "\n",
      "                 97.76% accuracy                   – 00:00:55 – 63.09 KB\n",
      "[350,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.71%   32.70%   45.49%     10063  \n",
      "       1      98.04%   99.67%   98.85%    339936  \n",
      "                                                  \n",
      "   Macro      86.38%   66.19%   72.17%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      97.37%   97.75%   97.32%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:01:04 – 64.64 KB\n",
      "[400,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.62%   32.88%   45.64%     11501  \n",
      "       1      98.05%   99.67%   98.85%    388498  \n",
      "                                                  \n",
      "   Macro      86.33%   66.27%   72.25%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      97.37%   97.75%   97.32%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:01:13 – 64.64 KB\n",
      "[450,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.26%   32.77%   45.47%     12936  \n",
      "       1      98.04%   99.66%   98.85%    437063  \n",
      "                                                  \n",
      "   Macro      86.15%   66.22%   72.16%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      97.36%   97.74%   97.31%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:01:22 – 63.09 KB\n",
      "[484,514]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      74.32%   32.69%   45.41%     13859  \n",
      "       1      98.05%   99.67%   98.85%    470654  \n",
      "                                                  \n",
      "   Macro      86.18%   66.18%   72.13%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      97.37%   97.75%   97.32%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:01:28 – 64.64 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      74.32%   32.69%   45.41%     13859  \n",
       "       1      98.05%   99.67%   98.85%    470654  \n",
       "                                                  \n",
       "   Macro      86.18%   66.18%   72.13%            \n",
       "   Micro      97.75%   97.75%   97.75%            \n",
       "Weighted      97.37%   97.75%   97.32%            \n",
       "\n",
       "                 97.75% accuracy                  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT-Recon.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model_bot_recon = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model_bot_recon, report, print_every=50000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "debad77b-302e-4482-b5e4-3bc55beb1e01",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save\n",
    "with open('model_bot_recon_con.pkl', 'wb') as f:\n",
    "    pickle.dump(model_bot_recon, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "1a714872-b73a-4b16-b816-77012f1dab0a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[5,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      98.96%   99.25%   99.11%      4417  \n",
      "       1      94.20%   92.10%   93.14%       582  \n",
      "                                                  \n",
      "   Macro      96.58%   95.67%   96.12%            \n",
      "   Micro      98.42%   98.42%   98.42%            \n",
      "Weighted      98.41%   98.42%   98.41%            \n",
      "\n",
      "                 98.42% accuracy                   – 00:00:00 – 57.27 KB\n",
      "[10,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      98.97%   99.39%   99.18%      8803  \n",
      "       1      95.34%   92.39%   93.84%      1196  \n",
      "                                                  \n",
      "   Macro      97.16%   95.89%   96.51%            \n",
      "   Micro      98.55%   98.55%   98.55%            \n",
      "Weighted      98.54%   98.55%   98.54%            \n",
      "\n",
      "                 98.55% accuracy                   – 00:00:01 – 60.08 KB\n",
      "[15,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.10%   99.37%   99.23%     13176  \n",
      "       1      95.36%   93.47%   94.40%      1823  \n",
      "                                                  \n",
      "   Macro      97.23%   96.42%   96.82%            \n",
      "   Micro      98.65%   98.65%   98.65%            \n",
      "Weighted      98.64%   98.65%   98.65%            \n",
      "\n",
      "                 98.65% accuracy                   – 00:00:02 – 61.34 KB\n",
      "[15,768]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.12%   99.37%   99.25%     13858  \n",
      "       1      95.36%   93.61%   94.48%      1909  \n",
      "                                                  \n",
      "   Macro      97.24%   96.49%   96.86%            \n",
      "   Micro      98.67%   98.67%   98.67%            \n",
      "Weighted      98.67%   98.67%   98.67%            \n",
      "\n",
      "                 98.67% accuracy                   – 00:00:02 – 61.34 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      99.12%   99.37%   99.25%     13858  \n",
       "       1      95.36%   93.61%   94.48%      1909  \n",
       "                                                  \n",
       "   Macro      97.24%   96.49%   96.86%            \n",
       "   Micro      98.67%   98.67%   98.67%            \n",
       "Weighted      98.67%   98.67%   98.67%            \n",
       "\n",
       "                 98.67% accuracy                  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT-Theft.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model_bot_theft = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model_bot_theft, report, print_every=5000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "a861278c-8f18-42fc-84b0-b5ae6f07f6de",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save\n",
    "with open('model_bot_theft_con.pkl', 'wb') as f:\n",
    "    pickle.dump(model_bot_theft, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26ec3afa-da59-4b94-9e3b-54d60a2dab3e",
   "metadata": {},
   "source": [
    "### Performance of NF BoT IoT dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a9caad1-4b4a-4ad3-af3b-0e6e2fa2b4a7",
   "metadata": {},
   "source": [
    "#### Without LTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "dec1c54c-70e4-420d-8719-5df2f81edd01",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[100,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      73.31%   27.45%   39.95%      2251  \n",
      "       1      98.35%   99.77%   99.06%     97748  \n",
      "                                                  \n",
      "   Macro      85.83%   63.61%   69.50%            \n",
      "   Micro      98.14%   98.14%   98.14%            \n",
      "Weighted      97.79%   98.14%   97.73%            \n",
      "\n",
      "                 98.14% accuracy                   – 00:00:17 – 72.52 KB\n",
      "[200,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      73.38%   27.83%   40.36%      4556  \n",
      "       1      98.34%   99.76%   99.05%    195443  \n",
      "                                                  \n",
      "   Macro      85.86%   63.80%   69.70%            \n",
      "   Micro      98.13%   98.13%   98.13%            \n",
      "Weighted      97.77%   98.13%   97.71%            \n",
      "\n",
      "                 98.13% accuracy                   – 00:00:35 – 74.06 KB\n",
      "[300,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.06%   28.09%   40.42%      6885  \n",
      "       1      98.33%   99.74%   99.03%    293114  \n",
      "                                                  \n",
      "   Macro      85.20%   63.92%   69.73%            \n",
      "   Micro      98.10%   98.10%   98.10%            \n",
      "Weighted      97.73%   98.10%   97.69%            \n",
      "\n",
      "                 98.10% accuracy                   – 00:00:54 – 72.52 KB\n",
      "[400,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      71.28%   27.73%   39.93%      9083  \n",
      "       1      98.34%   99.74%   99.04%    390916  \n",
      "                                                  \n",
      "   Macro      84.81%   63.74%   69.48%            \n",
      "   Micro      98.11%   98.11%   98.11%            \n",
      "Weighted      97.73%   98.11%   97.70%            \n",
      "\n",
      "                 98.11% accuracy                   – 00:01:12 – 74.06 KB\n",
      "[500,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      71.85%   27.82%   40.11%     11435  \n",
      "       1      98.33%   99.74%   99.03%    488564  \n",
      "                                                  \n",
      "   Macro      85.09%   63.78%   69.57%            \n",
      "   Micro      98.10%   98.10%   98.10%            \n",
      "Weighted      97.73%   98.10%   97.69%            \n",
      "\n",
      "                 98.10% accuracy                   – 00:01:30 – 72.52 KB\n",
      "[600,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.21%   28.09%   40.44%     13857  \n",
      "       1      98.32%   99.74%   99.03%    586142  \n",
      "                                                  \n",
      "   Macro      85.27%   63.92%   69.74%            \n",
      "   Micro      98.09%   98.09%   98.09%            \n",
      "Weighted      97.72%   98.09%   97.68%            \n",
      "\n",
      "                 98.09% accuracy                   – 00:01:48 – 74.06 KB\n",
      "[600,100]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.21%   28.09%   40.45%     13859  \n",
      "       1      98.32%   99.74%   99.03%    586240  \n",
      "                                                  \n",
      "   Macro      85.27%   63.92%   69.74%            \n",
      "   Micro      98.09%   98.09%   98.09%            \n",
      "Weighted      97.72%   98.09%   97.68%            \n",
      "\n",
      "                 98.09% accuracy                   – 00:01:48 – 74.06 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      72.21%   28.09%   40.45%     13859  \n",
       "       1      98.32%   99.74%   99.03%    586240  \n",
       "                                                  \n",
       "   Macro      85.27%   63.92%   69.74%            \n",
       "   Micro      98.09%   98.09%   98.09%            \n",
       "Weighted      97.72%   98.09%   97.68%            \n",
       "\n",
       "                 98.09% accuracy                  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model= compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=False, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model, report, print_every=100000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a1ba86f3-1c14-48d3-9138-697572448cb1",
   "metadata": {},
   "source": [
    "#### With LTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "7c10c400-e673-44a9-bfc3-3467c70a20e3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[100,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      73.31%   27.45%   39.95%      2251  \n",
      "       1      98.35%   99.77%   99.06%     97748  \n",
      "                                                  \n",
      "   Macro      85.83%   63.61%   69.50%            \n",
      "   Micro      98.14%   98.14%   98.14%            \n",
      "Weighted      97.79%   98.14%   97.73%            \n",
      "\n",
      "                 98.14% accuracy                   – 00:00:17 – 63.09 KB\n",
      "[200,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      73.38%   27.83%   40.36%      4556  \n",
      "       1      98.34%   99.76%   99.05%    195443  \n",
      "                                                  \n",
      "   Macro      85.86%   63.80%   69.70%            \n",
      "   Micro      98.13%   98.13%   98.13%            \n",
      "Weighted      97.77%   98.13%   97.71%            \n",
      "\n",
      "                 98.13% accuracy                   – 00:00:36 – 64.64 KB\n",
      "[300,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.06%   28.09%   40.42%      6885  \n",
      "       1      98.33%   99.74%   99.03%    293114  \n",
      "                                                  \n",
      "   Macro      85.20%   63.92%   69.73%            \n",
      "   Micro      98.10%   98.10%   98.10%            \n",
      "Weighted      97.73%   98.10%   97.69%            \n",
      "\n",
      "                 98.10% accuracy                   – 00:00:53 – 63.09 KB\n",
      "[400,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      71.28%   27.73%   39.93%      9083  \n",
      "       1      98.34%   99.74%   99.04%    390916  \n",
      "                                                  \n",
      "   Macro      84.81%   63.74%   69.48%            \n",
      "   Micro      98.11%   98.11%   98.11%            \n",
      "Weighted      97.73%   98.11%   97.70%            \n",
      "\n",
      "                 98.11% accuracy                   – 00:01:12 – 64.64 KB\n",
      "[500,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      71.85%   27.82%   40.11%     11435  \n",
      "       1      98.33%   99.74%   99.03%    488564  \n",
      "                                                  \n",
      "   Macro      85.09%   63.78%   69.57%            \n",
      "   Micro      98.10%   98.10%   98.10%            \n",
      "Weighted      97.73%   98.10%   97.69%            \n",
      "\n",
      "                 98.10% accuracy                   – 00:01:29 – 63.09 KB\n",
      "[600,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.21%   28.09%   40.44%     13857  \n",
      "       1      98.32%   99.74%   99.03%    586142  \n",
      "                                                  \n",
      "   Macro      85.27%   63.92%   69.74%            \n",
      "   Micro      98.09%   98.09%   98.09%            \n",
      "Weighted      97.72%   98.09%   97.68%            \n",
      "\n",
      "                 98.09% accuracy                   – 00:01:47 – 64.64 KB\n",
      "[600,100]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      72.21%   28.09%   40.45%     13859  \n",
      "       1      98.32%   99.74%   99.03%    586240  \n",
      "                                                  \n",
      "   Macro      85.27%   63.92%   69.74%            \n",
      "   Micro      98.09%   98.09%   98.09%            \n",
      "Weighted      97.72%   98.09%   97.68%            \n",
      "\n",
      "                 98.09% accuracy                   – 00:01:47 – 64.64 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      72.21%   28.09%   40.45%     13859  \n",
       "       1      98.32%   99.74%   99.03%    586240  \n",
       "                                                  \n",
       "   Macro      85.27%   63.92%   69.74%            \n",
       "   Micro      98.09%   98.09%   98.09%            \n",
       "Weighted      97.72%   98.09%   97.68%            \n",
       "\n",
       "                 98.09% accuracy                  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "model_bot_con = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "# Evaluate classifiers\n",
    "evaluate.progressive_val_score(dataset, model_bot_con, report, print_every=100000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "628191d1-505b-4009-9432-c8174398c237",
   "metadata": {},
   "source": [
    "### Saving the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "2e6545cf-ac34-43c4-a989-66d601281e23",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save\n",
    "with open('model_bot_con.pkl', 'wb') as f:\n",
    "    pickle.dump(model_bot_con, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f1b55d95-2e1f-41be-b211-1cb4841641c7",
   "metadata": {},
   "source": [
    "# Zero-Day Attack Detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "454543fb-f24c-42ef-9ad8-71d4365d6ef8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load \n",
    "with open('model_bot_con.pkl', 'rb') as f:\n",
    "    model_bot_con = pickle.load(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "40469e41-1f20-45e5-b86c-fcfd111ff710",
   "metadata": {},
   "source": [
    "#### Testing the DDoS model on BoT DDoS Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "46eaa633-fcd7-400a-b0a6-e137e071335b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load \n",
    "with open('model_bot_ddos_con.pkl', 'rb') as f:\n",
    "    model_bot_ddos = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "96b2541c-f368-4799-9f2c-73e636b43c73",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for Zero NF-BoT DDoS samples\n",
      "Classification Report:            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.54%   95.32%   97.38%     13859  \n",
      "       1      98.89%   99.89%   99.39%     57844  \n",
      "                                                  \n",
      "   Macro      99.21%   97.61%   98.39%            \n",
      "   Micro      99.01%   99.01%   99.01%            \n",
      "Weighted      99.02%   99.01%   99.00%            \n",
      "\n",
      "                 99.01% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-DDoS.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_ddos.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for Zero NF-BoT DDoS samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "814fba69-1f4f-4d1c-870c-68650015448b",
   "metadata": {},
   "source": [
    "#### Testing the BoT model on BoT DDoS Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "1ef8414a-a8a3-455a-877d-b430e5f8cb47",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT DDoS samples\n",
      "Classification Report:            Precision   Recall    F1       Support  \n",
      "                                                   \n",
      "       0     100.00%    31.72%   48.16%     13859  \n",
      "       1      85.94%   100.00%   92.44%     57844  \n",
      "                                                   \n",
      "   Macro      92.97%    65.86%   70.30%            \n",
      "   Micro      86.80%    86.80%   86.80%            \n",
      "Weighted      88.66%    86.80%   83.88%            \n",
      "\n",
      "                  86.80% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# # Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-DDoS.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_con.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT DDoS samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f6d1cb4-ae92-4253-8e5e-0fdc48949a5a",
   "metadata": {},
   "source": [
    "### Testing the DoS model on BoT DoS Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "72f9f6f2-ee65-40f6-8174-988bcf37640e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load \n",
    "with open('model_bot_dos_con.pkl', 'rb') as f:\n",
    "    model_bot_dos = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "6e0db158-f8bc-47ca-8f96-c4288be806ef",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT DoS samples\n",
      "Classification Report:            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      99.62%   93.58%   96.50%     13859  \n",
      "       1      98.48%   99.91%   99.19%     57833  \n",
      "                                                  \n",
      "   Macro      99.05%   96.75%   97.85%            \n",
      "   Micro      98.69%   98.69%   98.69%            \n",
      "Weighted      98.70%   98.69%   98.67%            \n",
      "\n",
      "                 98.69% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-DoS.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_dos.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT DoS samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1564359c-80f6-47de-9e84-29cfc82493c6",
   "metadata": {},
   "source": [
    "### Testing the BoT model on BoT DoS Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "b3cc967e-dede-4059-8b76-72077387875b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT DoS samples\n",
      "Classification Report:            Precision   Recall    F1       Support  \n",
      "                                                   \n",
      "       0     100.00%    31.72%   48.16%     13859  \n",
      "       1      85.94%   100.00%   92.44%     57833  \n",
      "                                                   \n",
      "   Macro      92.97%    65.86%   70.30%            \n",
      "   Micro      86.80%    86.80%   86.80%            \n",
      "Weighted      88.66%    86.80%   83.88%            \n",
      "\n",
      "                  86.80% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-DoS.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_con.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT DoS samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "29dcda48-801f-4de9-841f-279f0778a764",
   "metadata": {},
   "source": [
    "#### Testing the Reconnaissance model on BoT Reconnaissance Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b5d7d1e1-eb5b-49ba-b7ab-4aaec0d6632a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load \n",
    "with open('model_bot_recon_con.pkl', 'rb') as f:\n",
    "    model_bot_recon = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "bda4c047-8702-4a66-8edf-ecd570ebb5ff",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT Reconnaissance samples\n",
      "Classification Report:            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      96.78%   39.97%   56.57%     13859  \n",
      "       1      98.27%   99.96%   99.11%    471655  \n",
      "                                                  \n",
      "   Macro      97.53%   69.96%   77.84%            \n",
      "   Micro      98.25%   98.25%   98.25%            \n",
      "Weighted      98.22%   98.25%   97.89%            \n",
      "\n",
      "                 98.25% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-Recon.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_recon.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT Reconnaissance samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "67c0d199-43ef-4bb8-a567-2281246a9067",
   "metadata": {},
   "source": [
    "#### Testing the BoT model on BoT Reconnaissance Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "c42604d7-2255-4c85-9872-2bca9becd2ab",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT Reconnaissance samples\n",
      "Classification Report:            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      98.21%   31.72%   47.95%     13859  \n",
      "       1      98.03%   99.98%   99.00%    471655  \n",
      "                                                  \n",
      "   Macro      98.12%   65.85%   73.48%            \n",
      "   Micro      98.03%   98.03%   98.03%            \n",
      "Weighted      98.04%   98.03%   97.54%            \n",
      "\n",
      "                 98.03% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-Recon.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_con.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT Reconnaissance samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7f025e27-0756-4092-af17-c85eda413116",
   "metadata": {},
   "source": [
    "### Testing the Theft model on BoT Theft Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "d389bab8-8914-40e6-a548-b5c604385106",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load \n",
    "with open('model_bot_theft_con.pkl', 'rb') as f:\n",
    "    model_bot_theft = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "9e172fba-b428-4de9-be57-3079ed35bd7a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT Theft samples\n",
      "Classification Report:            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      97.21%   99.24%   98.21%     13859  \n",
      "       1      95.99%   86.42%   90.96%      2909  \n",
      "                                                  \n",
      "   Macro      96.60%   92.83%   94.58%            \n",
      "   Micro      97.02%   97.02%   97.02%            \n",
      "Weighted      97.00%   97.02%   96.96%            \n",
      "\n",
      "                 97.02% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-Theft.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_theft.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT Theft samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86ee7b14-f173-4f75-b7b2-43c408db15e6",
   "metadata": {},
   "source": [
    "### Testing the BoT model on BoT Theft Zero Day Attack"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "770245c8-a57a-4d3d-a90a-57fb043a11dd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Printing the results for FGSM NF-BoT Theft samples\n",
      "Classification Report:            Precision   Recall    F1       Support  \n",
      "                                                   \n",
      "       0     100.00%    31.72%   48.16%     13859  \n",
      "       1      23.51%   100.00%   38.07%      2909  \n",
      "                                                   \n",
      "   Macro      61.76%    65.86%   43.12%            \n",
      "   Micro      43.57%    43.57%   43.57%            \n",
      "Weighted      86.73%    43.57%   46.41%            \n",
      "\n",
      "                  43.57% accuracy                  \n"
     ]
    }
   ],
   "source": [
    "from river.metrics import ClassificationReport\n",
    "\n",
    "# Load datasets\n",
    "data = pd.read_csv(\"Augmented_NF-BoT-IoT-Theft.csv\")\n",
    "\n",
    "X = data.drop(['Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "\n",
    "# Initialize metrics\n",
    "Report = metrics.ClassificationReport()\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for xi, yi in stream.iter_pandas(X, y):\n",
    "    y_pred = model_bot_con.predict_one(xi)\n",
    "    Report.update(yi, y_pred)\n",
    "\n",
    "\n",
    "# Output the results\n",
    "print(\"Printing the results for FGSM NF-BoT Theft samples\")\n",
    "print(\"Classification Report:\", Report)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c1705930-c9f2-4c69-8ce5-95487643a44e",
   "metadata": {},
   "source": [
    "# Performance of SAMKNN "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "163e1443-a2bb-4e90-97bc-4d63122a86ec",
   "metadata": {},
   "source": [
    "### SAMKNN with LTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "57abcbf4-eb7f-4dd3-aa4e-42fa32863ff2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[50,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      53.53%    8.21%   14.24%      1108  \n",
      "       1      97.96%   99.84%   98.89%     48892  \n",
      "                                                  \n",
      "   Macro      75.74%   54.03%   56.57%            \n",
      "   Micro      97.81%   97.81%   97.81%            \n",
      "Weighted      96.97%   97.81%   97.01%            \n",
      "\n",
      "                 97.81% accuracy                   – 00:00:16 – 4.25 MB\n",
      "[100,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.95%    8.88%   15.30%      2251  \n",
      "       1      97.94%   99.83%   98.88%     97749  \n",
      "                                                  \n",
      "   Macro      76.44%   54.36%   57.09%            \n",
      "   Micro      97.78%   97.78%   97.78%            \n",
      "Weighted      96.97%   97.78%   97.00%            \n",
      "\n",
      "                 97.78% accuracy                   – 00:00:34 – 6.78 MB\n",
      "[150,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      53.82%    8.96%   15.37%      3381  \n",
      "       1      97.94%   99.82%   98.87%    146619  \n",
      "                                                  \n",
      "   Macro      75.88%   54.39%   57.12%            \n",
      "   Micro      97.77%   97.77%   97.77%            \n",
      "Weighted      96.95%   97.77%   96.99%            \n",
      "\n",
      "                 97.77% accuracy                   – 00:00:50 – 9.71 MB\n",
      "[200,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      53.60%    8.98%   15.38%      4556  \n",
      "       1      97.92%   99.82%   98.86%    195444  \n",
      "                                                  \n",
      "   Macro      75.76%   54.40%   57.12%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      96.91%   97.75%   96.96%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:01:07 – 12.21 MB\n",
      "[250,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.14%    9.01%   15.45%      5737  \n",
      "       1      97.90%   99.82%   98.85%    244263  \n",
      "                                                  \n",
      "   Macro      76.02%   54.42%   57.15%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      96.90%   97.74%   96.94%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:01:23 – 14.99 MB\n",
      "[300,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.34%    8.92%   15.32%      6885  \n",
      "       1      97.90%   99.82%   98.85%    293115  \n",
      "                                                  \n",
      "   Macro      76.12%   54.37%   57.09%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      96.90%   97.74%   96.94%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:01:39 – 18.08 MB\n",
      "[350,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      55.17%    9.18%   15.74%      7962  \n",
      "       1      97.93%   99.83%   98.87%    342038  \n",
      "                                                  \n",
      "   Macro      76.55%   54.50%   57.30%            \n",
      "   Micro      97.76%   97.76%   97.76%            \n",
      "Weighted      96.95%   97.76%   96.98%            \n",
      "\n",
      "                 97.76% accuracy                   – 00:01:56 – 20.54 MB\n",
      "[400,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.96%    8.91%   15.33%      9083  \n",
      "       1      97.92%   99.83%   98.87%    390917  \n",
      "                                                  \n",
      "   Macro      76.44%   54.37%   57.10%            \n",
      "   Micro      97.77%   97.77%   97.77%            \n",
      "Weighted      96.95%   97.77%   96.97%            \n",
      "\n",
      "                 97.77% accuracy                   – 00:02:13 – 23.11 MB\n",
      "[450,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      55.34%    8.99%   15.47%     10253  \n",
      "       1      97.92%   99.83%   98.87%    439747  \n",
      "                                                  \n",
      "   Macro      76.63%   54.41%   57.17%            \n",
      "   Micro      97.76%   97.76%   97.76%            \n",
      "Weighted      96.95%   97.76%   96.97%            \n",
      "\n",
      "                 97.76% accuracy                   – 00:02:29 – 25.82 MB\n",
      "[500,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.69%    8.76%   15.11%     11435  \n",
      "       1      97.91%   99.83%   98.86%    488565  \n",
      "                                                  \n",
      "   Macro      76.30%   54.30%   56.98%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      96.92%   97.75%   96.94%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:02:45 – 28.68 MB\n",
      "[550,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      55.08%    8.80%   15.18%     12676  \n",
      "       1      97.89%   99.83%   98.85%    537324  \n",
      "                                                  \n",
      "   Macro      76.49%   54.32%   57.02%            \n",
      "   Micro      97.73%   97.73%   97.73%            \n",
      "Weighted      96.90%   97.73%   96.92%            \n",
      "\n",
      "                 97.73% accuracy                   – 00:03:01 – 31.68 MB\n",
      "[600,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.84%    8.79%   15.15%     13857  \n",
      "       1      97.89%   99.83%   98.85%    586143  \n",
      "                                                  \n",
      "   Macro      76.36%   54.31%   57.00%            \n",
      "   Micro      97.73%   97.73%   97.73%            \n",
      "Weighted      96.89%   97.73%   96.91%            \n",
      "\n",
      "                 97.73% accuracy                   – 00:03:18 – 34.9 MB\n",
      "[600,100]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      54.84%    8.79%   15.15%     13859  \n",
      "       1      97.89%   99.83%   98.85%    586241  \n",
      "                                                  \n",
      "   Macro      76.36%   54.31%   57.00%            \n",
      "   Micro      97.73%   97.73%   97.73%            \n",
      "Weighted      96.89%   97.73%   96.91%            \n",
      "\n",
      "                 97.73% accuracy                   – 00:03:18 – 34.92 MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      54.84%    8.79%   15.15%     13859  \n",
       "       1      97.89%   99.83%   98.85%    586241  \n",
       "                                                  \n",
       "   Macro      76.36%   54.31%   57.00%            \n",
       "   Micro      97.73%   97.73%   97.73%            \n",
       "Weighted      96.89%   97.73%   96.91%            \n",
       "\n",
       "                 97.73% accuracy                  "
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "SAMKNN_model = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    SAMKNNClassifier(ltm_size=0.4, max_window_size=300, min_stm_size=30, n_neighbors=5, stm_size_option='maxACCApprox', use_ltm=True, weighting='distance')\n",
    ")\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "evaluate.progressive_val_score(dataset, SAMKNN_model, report, print_every=50000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "78bf2df8-4f98-4a76-b055-082184eb58fd",
   "metadata": {},
   "source": [
    "### SAMKNN without LTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "c9b9f362-64ac-4ecf-8704-1d2991d51f7e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[50,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      48.94%    8.30%   14.20%      1108  \n",
      "       1      97.96%   99.80%   98.87%     48892  \n",
      "                                                  \n",
      "   Macro      73.45%   54.05%   56.54%            \n",
      "   Micro      97.78%   97.78%   97.78%            \n",
      "Weighted      96.87%   97.78%   97.00%            \n",
      "\n",
      "                 97.78% accuracy                   – 00:00:06 – 4.71 MB\n",
      "[100,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.16%    9.77%   16.41%      2251  \n",
      "       1      97.96%   99.79%   98.86%     97749  \n",
      "                                                  \n",
      "   Macro      74.56%   54.78%   57.64%            \n",
      "   Micro      97.76%   97.76%   97.76%            \n",
      "Weighted      96.91%   97.76%   97.01%            \n",
      "\n",
      "                 97.76% accuracy                   – 00:00:12 – 5.42 MB\n",
      "[150,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      50.31%    9.73%   16.31%      3381  \n",
      "       1      97.96%   99.78%   98.86%    146619  \n",
      "                                                  \n",
      "   Macro      74.13%   54.75%   57.58%            \n",
      "   Micro      97.75%   97.75%   97.75%            \n",
      "Weighted      96.88%   97.75%   97.00%            \n",
      "\n",
      "                 97.75% accuracy                   – 00:00:17 – 6.38 MB\n",
      "[200,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      49.83%    9.75%   16.30%      4556  \n",
      "       1      97.93%   99.77%   98.84%    195444  \n",
      "                                                  \n",
      "   Macro      73.88%   54.76%   57.57%            \n",
      "   Micro      97.72%   97.72%   97.72%            \n",
      "Weighted      96.84%   97.72%   96.96%            \n",
      "\n",
      "                 97.72% accuracy                   – 00:00:23 – 7.03 MB\n",
      "[250,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      50.09%    9.69%   16.24%      5737  \n",
      "       1      97.92%   99.77%   98.84%    244263  \n",
      "                                                  \n",
      "   Macro      74.00%   54.73%   57.54%            \n",
      "   Micro      97.71%   97.71%   97.71%            \n",
      "Weighted      96.82%   97.71%   96.94%            \n",
      "\n",
      "                 97.71% accuracy                   – 00:00:28 – 7.88 MB\n",
      "[300,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      50.80%    9.73%   16.33%      6885  \n",
      "       1      97.92%   99.78%   98.84%    293115  \n",
      "                                                  \n",
      "   Macro      74.36%   54.75%   57.59%            \n",
      "   Micro      97.71%   97.71%   97.71%            \n",
      "Weighted      96.84%   97.71%   96.95%            \n",
      "\n",
      "                 97.71% accuracy                   – 00:00:33 – 8.93 MB\n",
      "[350,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.73%    9.97%   16.72%      7962  \n",
      "       1      97.94%   99.78%   98.85%    342038  \n",
      "                                                  \n",
      "   Macro      74.83%   54.88%   57.79%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      96.89%   97.74%   96.99%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:00:39 – 9.59 MB\n",
      "[400,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.77%    9.80%   16.48%      9083  \n",
      "       1      97.94%   99.79%   98.86%    390917  \n",
      "                                                  \n",
      "   Macro      74.86%   54.79%   57.67%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      96.89%   97.74%   96.99%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:00:44 – 10.32 MB\n",
      "[450,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      52.09%    9.83%   16.54%     10253  \n",
      "       1      97.94%   99.79%   98.85%    439747  \n",
      "                                                  \n",
      "   Macro      75.01%   54.81%   57.70%            \n",
      "   Micro      97.74%   97.74%   97.74%            \n",
      "Weighted      96.89%   97.74%   96.98%            \n",
      "\n",
      "                 97.74% accuracy                   – 00:00:50 – 11.12 MB\n",
      "[500,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.45%    9.62%   16.21%     11435  \n",
      "       1      97.92%   99.79%   98.85%    488565  \n",
      "                                                  \n",
      "   Macro      74.69%   54.70%   57.53%            \n",
      "   Micro      97.73%   97.73%   97.73%            \n",
      "Weighted      96.86%   97.73%   96.96%            \n",
      "\n",
      "                 97.73% accuracy                   – 00:00:55 – 12.02 MB\n",
      "[550,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.74%    9.71%   16.35%     12676  \n",
      "       1      97.91%   99.79%   98.84%    537324  \n",
      "                                                  \n",
      "   Macro      74.83%   54.75%   57.60%            \n",
      "   Micro      97.71%   97.71%   97.71%            \n",
      "Weighted      96.85%   97.71%   96.94%            \n",
      "\n",
      "                 97.71% accuracy                   – 00:01:00 – 13.03 MB\n",
      "[600,000]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.51%    9.71%   16.33%     13857  \n",
      "       1      97.91%   99.78%   98.84%    586143  \n",
      "                                                  \n",
      "   Macro      74.71%   54.75%   57.59%            \n",
      "   Micro      97.70%   97.70%   97.70%            \n",
      "Weighted      96.83%   97.70%   96.93%            \n",
      "\n",
      "                 97.70% accuracy                   – 00:01:06 – 14.18 MB\n",
      "[600,100]            Precision   Recall   F1       Support  \n",
      "                                                  \n",
      "       0      51.51%    9.70%   16.33%     13859  \n",
      "       1      97.91%   99.78%   98.84%    586241  \n",
      "                                                  \n",
      "   Macro      74.71%   54.74%   57.58%            \n",
      "   Micro      97.70%   97.70%   97.70%            \n",
      "Weighted      96.83%   97.70%   96.93%            \n",
      "\n",
      "                 97.70% accuracy                   – 00:01:06 – 14.17 MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "           Precision   Recall   F1       Support  \n",
       "                                                  \n",
       "       0      51.51%    9.70%   16.33%     13859  \n",
       "       1      97.91%   99.78%   98.84%    586241  \n",
       "                                                  \n",
       "   Macro      74.71%   54.74%   57.58%            \n",
       "   Micro      97.70%   97.70%   97.70%            \n",
       "Weighted      96.83%   97.70%   96.93%            \n",
       "\n",
       "                 97.70% accuracy                  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "dataset = stream.iter_pandas(X, y)\n",
    "\n",
    "SAMKNN_model = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    SAMKNNClassifier(ltm_size=0.4, max_window_size=500, min_stm_size=30, n_neighbors=5, stm_size_option='maxACCApprox', use_ltm=False, weighting='distance')\n",
    ")\n",
    "#Passing the evalaution metrics into a variable\n",
    "report = metrics.ClassificationReport()\n",
    "\n",
    "evaluate.progressive_val_score(dataset, SAMKNN_model, report, print_every=50000, show_time=True, show_memory=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "46bef191-708b-48ac-9ef7-4bc255fc30be",
   "metadata": {},
   "source": [
    "### Statistical Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "57a301cf-e820-4634-a8a5-c6ad63915d73",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data sampled to 100000 instances for faster processing.\n",
      "Class Distribution:\n",
      "Label\n",
      "1    97749\n",
      "0     2251\n",
      "Name: count, dtype: int64\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs:   0%|                                               | 0/5 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Run 1/5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs:  20%|███████▊                               | 1/5 [00:48<03:12, 48.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaptiveSAMKNNClassifier Accuracy: 0.9813\n",
      "SAMKNNClassifier Accuracy: 0.9779\n",
      "\n",
      "Run 2/5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs:  40%|███████████████▌                       | 2/5 [01:36<02:25, 48.51s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaptiveSAMKNNClassifier Accuracy: 0.9818\n",
      "SAMKNNClassifier Accuracy: 0.9781\n",
      "\n",
      "Run 3/5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs:  60%|███████████████████████▍               | 3/5 [02:24<01:36, 48.12s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaptiveSAMKNNClassifier Accuracy: 0.9816\n",
      "SAMKNNClassifier Accuracy: 0.9776\n",
      "\n",
      "Run 4/5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs:  80%|███████████████████████████████▏       | 4/5 [03:14<00:48, 48.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaptiveSAMKNNClassifier Accuracy: 0.9814\n",
      "SAMKNNClassifier Accuracy: 0.9775\n",
      "\n",
      "Run 5/5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Runs: 100%|███████████████████████████████████████| 5/5 [04:03<00:00, 48.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaptiveSAMKNNClassifier Accuracy: 0.9816\n",
      "SAMKNNClassifier Accuracy: 0.9778\n",
      "\n",
      "=== Performance Metrics ===\n",
      "AdaptiveSAMKNNClassifier Accuracy: [0.98132 0.98177 0.98158 0.98139 0.98158]\n",
      "Mean Accuracy: 0.9815\n",
      "Standard Deviation: 0.0002\n",
      "95% Confidence Interval: (0.9813, 0.9817)\n",
      "\n",
      "SAMKNNClassifier Accuracy: [0.97789 0.97806 0.97764 0.97749 0.97777]\n",
      "Mean Accuracy: 0.9778\n",
      "Standard Deviation: 0.0002\n",
      "95% Confidence Interval: (0.9775, 0.9780)\n",
      "\n",
      "=== Statistical Test ===\n",
      "Paired t-test: t-statistic = 41.2643, p-value = 0.0000\n",
      "Result: The difference in accuracy is statistically significant.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from sklearn.utils import shuffle\n",
    "from river import stream, compose, preprocessing, metrics\n",
    "import numpy as np\n",
    "from scipy import stats\n",
    "from tqdm import tqdm  # Ensure tqdm is imported\n",
    "\n",
    "# Load and Prepare Data\n",
    "data = pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "\n",
    "# Sample the data for faster testing \n",
    "sample_size = 100000  \n",
    "if len(data) > sample_size:\n",
    "    data = data.sample(n=sample_size, random_state=42)\n",
    "    print(f\"Data sampled to {sample_size} instances for faster processing.\")\n",
    "\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "\n",
    "# **Option 1:** Keep labels as integers\n",
    "# Ensure that labels are integers\n",
    "y = y.astype(int)\n",
    "\n",
    "# Check class distribution\n",
    "class_counts = y.value_counts()\n",
    "print(\"Class Distribution:\")\n",
    "print(class_counts)\n",
    "\n",
    "# Parameters for Evaluation\n",
    "n_runs = 5  # Number of independent evaluation runs\n",
    "random_states = np.random.randint(0, 10000, size=n_runs)\n",
    "\n",
    "# Containers to Store Performance Metrics\n",
    "accuracy_adaptive = []\n",
    "accuracy_samknn = []\n",
    "\n",
    "# Define Classifiers\n",
    "def get_models():\n",
    "    model_adaptive = compose.Pipeline(\n",
    "        preprocessing.StandardScaler(),\n",
    "        ('classifier', AdaptiveSAMKNNClassifier(\n",
    "            max_window_size=300,  \n",
    "            ltm_size=0.4,\n",
    "            use_ltm=True,\n",
    "            max_ltm_size_threshold=0.30\n",
    "        ))\n",
    "    )\n",
    "    \n",
    "    model_samknn = compose.Pipeline(\n",
    "        preprocessing.StandardScaler(),\n",
    "        ('classifier', SAMKNNClassifier(\n",
    "            ltm_size=0.4, \n",
    "            max_window_size=300, \n",
    "            min_stm_size=30, \n",
    "            n_neighbors=5, \n",
    "            stm_size_option='maxACCApprox', \n",
    "            use_ltm=True\n",
    "        ))\n",
    "    )\n",
    "    \n",
    "    return model_adaptive, model_samknn\n",
    "\n",
    "# Evaluation Loop with Optimizations\n",
    "for i in tqdm(range(n_runs), desc=\"Runs\"):\n",
    "    print(f\"\\nRun {i+1}/{n_runs}\")\n",
    "    \n",
    "    # Shuffle Data with a Unique Random State\n",
    "    X_shuffled, y_shuffled = shuffle(X, y, random_state=random_states[i])\n",
    "    dataset = stream.iter_pandas(X_shuffled, y_shuffled)\n",
    "    \n",
    "    # Initialize Models\n",
    "    model_adaptive, model_samknn = get_models()\n",
    "    \n",
    "    # Initialize Metrics\n",
    "    metric_adaptive = metrics.Accuracy()\n",
    "    metric_samknn = metrics.Accuracy()\n",
    "    \n",
    "    # Evaluate both classifiers in a single pass\n",
    "    for xi, yi_label in dataset:\n",
    "        # 'xi' is already a dictionary; no need to convert\n",
    "        \n",
    "        # AdaptiveSAMKNNClassifier\n",
    "        y_pred_adaptive = model_adaptive.predict_one(xi)\n",
    "        model_adaptive.learn_one(xi, yi_label)\n",
    "        metric_adaptive.update(yi_label, y_pred_adaptive)\n",
    "        \n",
    "        # SAMKNNClassifier\n",
    "        y_pred_knn = model_samknn.predict_one(xi)\n",
    "        model_samknn.learn_one(xi, yi_label)\n",
    "        metric_samknn.update(yi_label, y_pred_knn)\n",
    "        \n",
    "    # Collect Metrics for the Current Run\n",
    "    fold_accuracy_adaptive = metric_adaptive.get()\n",
    "    fold_accuracy_samknn = metric_samknn.get()  # Corrected line\n",
    "    \n",
    "    accuracy_adaptive.append(fold_accuracy_adaptive)\n",
    "    accuracy_samknn.append(fold_accuracy_samknn)\n",
    "    \n",
    "    # Display Metrics for the Current Run\n",
    "    print(f\"AdaptiveSAMKNNClassifier Accuracy: {fold_accuracy_adaptive:.4f}\")\n",
    "    print(f\"SAMKNNClassifier Accuracy: {fold_accuracy_samknn:.4f}\")\n",
    "\n",
    "# Statistical Testing\n",
    "\n",
    "# Convert lists to NumPy arrays for easier manipulation\n",
    "accuracy_adaptive = np.array(accuracy_adaptive)\n",
    "accuracy_samknn = np.array(accuracy_samknn)\n",
    "\n",
    "# Calculate Mean and Standard Deviation\n",
    "mean_adaptive = accuracy_adaptive.mean()\n",
    "std_adaptive = accuracy_adaptive.std(ddof=1)\n",
    "mean_samknn = accuracy_samknn.mean()\n",
    "std_samknn = accuracy_samknn.std(ddof=1)\n",
    "\n",
    "# Calculate 95% Confidence Intervals\n",
    "conf_int_adaptive = stats.t.interval(\n",
    "    0.95, \n",
    "    len(accuracy_adaptive)-1, \n",
    "    loc=mean_adaptive, \n",
    "    scale=std_adaptive/np.sqrt(n_runs)\n",
    ")\n",
    "\n",
    "conf_int_samknn = stats.t.interval(\n",
    "    0.95, \n",
    "    len(accuracy_samknn)-1, \n",
    "    loc=mean_samknn, \n",
    "    scale=std_samknn/np.sqrt(n_runs)\n",
    ")\n",
    "\n",
    "# Perform Paired t-test\n",
    "t_stat, p_value = stats.ttest_rel(accuracy_adaptive, accuracy_samknn)\n",
    "\n",
    "# Step 6: Display Results\n",
    "print(\"\\n=== Performance Metrics ===\")\n",
    "print(f\"AdaptiveSAMKNNClassifier Accuracy: {accuracy_adaptive}\")\n",
    "print(f\"Mean Accuracy: {mean_adaptive:.4f}\")\n",
    "print(f\"Standard Deviation: {std_adaptive:.4f}\")\n",
    "print(f\"95% Confidence Interval: ({conf_int_adaptive[0]:.4f}, {conf_int_adaptive[1]:.4f})\\n\")\n",
    "\n",
    "print(f\"SAMKNNClassifier Accuracy: {accuracy_samknn}\")\n",
    "print(f\"Mean Accuracy: {mean_samknn:.4f}\")\n",
    "print(f\"Standard Deviation: {std_samknn:.4f}\")\n",
    "print(f\"95% Confidence Interval: ({conf_int_samknn[0]:.4f}, {conf_int_samknn[1]:.4f})\\n\")\n",
    "\n",
    "print(\"=== Statistical Test ===\")\n",
    "print(f\"Paired t-test: t-statistic = {t_stat:.4f}, p-value = {p_value:.4f}\")\n",
    "\n",
    "if p_value < 0.05:\n",
    "    print(\"Result: The difference in accuracy is statistically significant.\")\n",
    "else:\n",
    "    print(\"Result: The difference in accuracy is not statistically significant.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "669276b5-876f-491a-954e-1026686d1f17",
   "metadata": {},
   "source": [
    "### Visualising the Performance of BoT IoT datset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "1d72a671-2465-4f40-8f19-86566163a13e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from collections import deque\n",
    "import numpy as np\n",
    "from river import metrics, stream, compose, preprocessing\n",
    "from sklearn.utils import shuffle\n",
    "import pandas as pd\n",
    "\n",
    "# Load dataset\n",
    "data = pd.read_csv(\"/Users/promisea/SAMKNN/NF-BoT-IoT/NF-BoT-IoT.csv\")\n",
    "\n",
    "# Prepare features and target\n",
    "X = data.drop(['IPV4_SRC_ADDR', 'IPV4_DST_ADDR', 'Attack', 'Label'], axis=1)\n",
    "y = data['Label']\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "# Initialize rolling window size\n",
    "window_size = 500\n",
    "\n",
    "# Initialize rolling metrics storage\n",
    "rolling_acc = deque(maxlen=window_size)\n",
    "rolling_pre = deque(maxlen=window_size)\n",
    "rolling_rec = deque(maxlen=window_size)\n",
    "rolling_f1 = deque(maxlen=window_size)\n",
    "\n",
    "# Initialize lists for storing the cumulative metrics for plotting\n",
    "acc_values = []\n",
    "pre_values = []\n",
    "rec_values = []\n",
    "f1_values = []\n",
    "window_indices = []\n",
    "\n",
    "# Model setup\n",
    "model = compose.Pipeline(\n",
    "    preprocessing.StandardScaler(),\n",
    "    ('classifier', AdaptiveSAMKNNClassifier(max_window_size=300, ltm_size=0.4, use_ltm=True, max_ltm_size_threshold=0.30))\n",
    ")\n",
    "\n",
    "# Initialize metrics\n",
    "Acc = metrics.Accuracy()\n",
    "Pre = metrics.Precision()\n",
    "Rec = metrics.Recall()\n",
    "F1 = metrics.F1()\n",
    "\n",
    "# Shuffle dataset\n",
    "X, y = shuffle(X, y, random_state=42)\n",
    "\n",
    "# Iterate over the test stream and update metrics\n",
    "for idx, (xi, yi) in enumerate(stream.iter_pandas(X, y)):\n",
    "    y_pred = model.predict_one(xi)\n",
    "    model.learn_one(xi, yi)\n",
    "    \n",
    "    # Update metrics for the current sample\n",
    "    Acc.update(yi, y_pred)\n",
    "    Pre.update(yi, y_pred)\n",
    "    Rec.update(yi, y_pred)\n",
    "    F1.update(yi, y_pred)\n",
    "    \n",
    "    # Update rolling metrics\n",
    "    rolling_acc.append(Acc.get())\n",
    "    rolling_pre.append(Pre.get())\n",
    "    rolling_rec.append(Rec.get())\n",
    "    rolling_f1.append(F1.get())\n",
    "    \n",
    "    # After every 500 samples, store the average of the rolling window\n",
    "    if len(rolling_acc) == window_size:\n",
    "        acc_values.append(np.mean(rolling_acc))\n",
    "        pre_values.append(np.mean(rolling_pre))\n",
    "        rec_values.append(np.mean(rolling_rec))\n",
    "        f1_values.append(np.mean(rolling_f1))\n",
    "        window_indices.append(idx)\n",
    "\n",
    "# Plot\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(window_indices, acc_values, label='Accuracy', color='blue')\n",
    "plt.plot(window_indices, pre_values, label='Precision', color='green')\n",
    "plt.plot(window_indices, rec_values, label='Recall', color='orange')\n",
    "plt.plot(window_indices, f1_values, label='F1 Score', color='red')\n",
    "\n",
    "# Make the title and labels bold\n",
    "plt.xlabel('Number of Samples', fontsize=12, fontweight='bold')\n",
    "plt.ylabel('Metric Value', fontsize=12, fontweight='bold')\n",
    "plt.title('The performance of the proposed model on NF BoT IoT dataset', fontsize=14, fontweight='bold')\n",
    "\n",
    "# Make the legend bold\n",
    "plt.legend(fontsize=10, title_fontsize='12', title='Metrics', loc='best')\n",
    "\n",
    "# Make the ticks bold\n",
    "plt.xticks(fontsize=10, fontweight='bold')\n",
    "plt.yticks(fontsize=10, fontweight='bold')\n",
    "\n",
    "plt.grid(True)\n",
    "plt.savefig('NF_BoT_Conference.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "415c7c94-b26e-46fc-bcae-a98bc1eedb9f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "river",
   "language": "python",
   "name": "river"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}