CICIOT2023.ipynb

{"cells":[{"cell_type":"code","execution_count":1,"metadata":{"executionInfo":{"elapsed":10978,"status":"ok","timestamp":1705010650990,"user":{"displayName":"Hiếu Võ","userId":"16914104909337499796"},"user_tz":-420},"id":"ci7lGMIlWDOi"},"outputs":[{"name":"stderr","output_type":"stream","text":["2024-03-01 06:24:48.566488: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.\n","2024-03-01 06:24:48.641715: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n","2024-03-01 06:24:48.641772: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n","2024-03-01 06:24:48.645411: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n","2024-03-01 06:24:48.659530: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.\n","2024-03-01 06:24:48.660476: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n","To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n","2024-03-01 06:24:50.457578: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"]}],"source":["import glob\n","import pandas as pd\n","import numpy as np\n","from numpy.linalg import norm\n","from sklearn.preprocessing import MinMaxScaler\n","from sklearn.model_selection import train_test_split\n","from tensorflow.keras.models import Sequential\n","from tensorflow.keras.layers import Dense\n","from tensorflow.keras.optimizers import SGD, Adam\n","from tensorflow.keras.layers import Dropout\n","import tensorflow as tf\n","from tqdm import tqdm\n","from tensorflow_privacy.privacy.optimizers.dp_optimizer_keras import DPKerasSGDOptimizer\n","\n","%matplotlib inline"]},{"cell_type":"code","execution_count":2,"metadata":{},"outputs":[],"source":["dict_2classes = {}\n","dict_2classes['DDoS-RSTFINFlood'] = 'Attack'\n","dict_2classes['DDoS-PSHACK_Flood'] = 'Attack'\n","dict_2classes['DDoS-SYN_Flood'] = 'Attack'\n","dict_2classes['DDoS-UDP_Flood'] = 'Attack'\n","dict_2classes['DDoS-TCP_Flood'] = 'Attack'\n","dict_2classes['DDoS-ICMP_Flood'] = 'Attack'\n","dict_2classes['DDoS-SynonymousIP_Flood'] = 'Attack'\n","dict_2classes['DDoS-ACK_Fragmentation'] = 'Attack'\n","dict_2classes['DDoS-UDP_Fragmentation'] = 'Attack'\n","dict_2classes['DDoS-ICMP_Fragmentation'] = 'Attack'\n","dict_2classes['DDoS-SlowLoris'] = 'Attack'\n","dict_2classes['DDoS-HTTP_Flood'] = 'Attack'\n","\n","dict_2classes['DoS-UDP_Flood'] = 'Attack'\n","dict_2classes['DoS-SYN_Flood'] = 'Attack'\n","dict_2classes['DoS-TCP_Flood'] = 'Attack'\n","dict_2classes['DoS-HTTP_Flood'] = 'Attack'\n","\n","\n","dict_2classes['Mirai-greeth_flood'] = 'Attack'\n","dict_2classes['Mirai-greip_flood'] = 'Attack'\n","dict_2classes['Mirai-udpplain'] = 'Attack'\n","\n","dict_2classes['Recon-PingSweep'] = 'Attack'\n","dict_2classes['Recon-OSScan'] = 'Attack'\n","dict_2classes['Recon-PortScan'] = 'Attack'\n","dict_2classes['VulnerabilityScan'] = 'Attack'\n","dict_2classes['Recon-HostDiscovery'] = 'Attack'\n","\n","dict_2classes['DNS_Spoofing'] = 'Attack'\n","dict_2classes['MITM-ArpSpoofing'] = 'Attack'\n","\n","dict_2classes['BenignTraffic'] = 'Benign'\n","\n","dict_2classes['BrowserHijacking'] = 'Attack'\n","dict_2classes['Backdoor_Malware'] = 'Attack'\n","dict_2classes['XSS'] = 'Attack'\n","dict_2classes['Uploading_Attack'] = 'Attack'\n","dict_2classes['SqlInjection'] = 'Attack'\n","dict_2classes['CommandInjection'] = 'Attack'\n","\n","dict_2classes['DictionaryBruteForce'] = 'Attack'"]},{"cell_type":"code","execution_count":3,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["file_paths len:  169\n"]},{"name":"stderr","output_type":"stream","text":["Processing files:   0%|          | 0/2 [00:00<?, ?file/s]"]},{"name":"stderr","output_type":"stream","text":["Processing files: 100%|██████████| 2/2 [00:02<00:00,  1.44s/file]\n"]}],"source":["\n","file_paths = glob.glob('/home/haohao/Downloads/archive/wataiData/csv/CICIoT2023/*.csv')\n","file_paths.sort()\n","print('file_paths len: ', len(file_paths))\n","\n","df = pd.DataFrame()\n","#[:round(len(file_paths)*0.01)]\n","for file_path in tqdm(file_paths[:round(len(file_paths)*0.01)], desc='Processing files', unit='file'):\n","    dff = pd.read_csv(file_path)\n","    \n","    df = pd.concat([df, dff], ignore_index=True)"]},{"cell_type":"code","execution_count":4,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["<class 'pandas.core.frame.DataFrame'>\n","RangeIndex: 457492 entries, 0 to 457491\n","Data columns (total 47 columns):\n"," #   Column           Non-Null Count   Dtype  \n","---  ------           --------------   -----  \n"," 0   flow_duration    457492 non-null  float64\n"," 1   Header_Length    457492 non-null  float64\n"," 2   Protocol Type    457492 non-null  float64\n"," 3   Duration         457492 non-null  float64\n"," 4   Rate             457492 non-null  float64\n"," 5   Srate            457492 non-null  float64\n"," 6   Drate            457492 non-null  float64\n"," 7   fin_flag_number  457492 non-null  float64\n"," 8   syn_flag_number  457492 non-null  float64\n"," 9   rst_flag_number  457492 non-null  float64\n"," 10  psh_flag_number  457492 non-null  float64\n"," 11  ack_flag_number  457492 non-null  float64\n"," 12  ece_flag_number  457492 non-null  float64\n"," 13  cwr_flag_number  457492 non-null  float64\n"," 14  ack_count        457492 non-null  float64\n"," 15  syn_count        457492 non-null  float64\n"," 16  fin_count        457492 non-null  float64\n"," 17  urg_count        457492 non-null  float64\n"," 18  rst_count        457492 non-null  float64\n"," 19  HTTP             457492 non-null  float64\n"," 20  HTTPS            457492 non-null  float64\n"," 21  DNS              457492 non-null  float64\n"," 22  Telnet           457492 non-null  float64\n"," 23  SMTP             457492 non-null  float64\n"," 24  SSH              457492 non-null  float64\n"," 25  IRC              457492 non-null  float64\n"," 26  TCP              457492 non-null  float64\n"," 27  UDP              457492 non-null  float64\n"," 28  DHCP             457492 non-null  float64\n"," 29  ARP              457492 non-null  float64\n"," 30  ICMP             457492 non-null  float64\n"," 31  IPv              457492 non-null  float64\n"," 32  LLC              457492 non-null  float64\n"," 33  Tot sum          457492 non-null  float64\n"," 34  Min              457492 non-null  float64\n"," 35  Max              457492 non-null  float64\n"," 36  AVG              457492 non-null  float64\n"," 37  Std              457492 non-null  float64\n"," 38  Tot size         457492 non-null  float64\n"," 39  IAT              457492 non-null  float64\n"," 40  Number           457492 non-null  float64\n"," 41  Magnitue         457492 non-null  float64\n"," 42  Radius           457492 non-null  float64\n"," 43  Covariance       457492 non-null  float64\n"," 44  Variance         457492 non-null  float64\n"," 45  Weight           457492 non-null  float64\n"," 46  label            457492 non-null  object \n","dtypes: float64(46), object(1)\n","memory usage: 164.0+ MB\n"]}],"source":["df.info()"]},{"cell_type":"code","execution_count":5,"metadata":{},"outputs":[],"source":["df['label'] = df['label'].map(dict_2classes)\n","df = df.sample(frac=1, random_state=42)"]},{"cell_type":"code","execution_count":6,"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>flow_duration</th>\n","      <th>Header_Length</th>\n","      <th>Protocol Type</th>\n","      <th>Duration</th>\n","      <th>Rate</th>\n","      <th>Srate</th>\n","      <th>Drate</th>\n","      <th>fin_flag_number</th>\n","      <th>syn_flag_number</th>\n","      <th>rst_flag_number</th>\n","      <th>...</th>\n","      <th>Std</th>\n","      <th>Tot size</th>\n","      <th>IAT</th>\n","      <th>Number</th>\n","      <th>Magnitue</th>\n","      <th>Radius</th>\n","      <th>Covariance</th>\n","      <th>Variance</th>\n","      <th>Weight</th>\n","      <th>label</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>413784</th>\n","      <td>0.130570</td>\n","      <td>30076.00</td>\n","      <td>17.00</td>\n","      <td>64.00</td>\n","      <td>4580.297292</td>\n","      <td>4580.297292</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>50.00</td>\n","      <td>8.310187e+07</td>\n","      <td>9.5</td>\n","      <td>10.000000</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>433656</th>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>1.00</td>\n","      <td>64.00</td>\n","      <td>37.205150</td>\n","      <td>37.205150</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>42.00</td>\n","      <td>8.314968e+07</td>\n","      <td>9.5</td>\n","      <td>9.165151</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>148341</th>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>6.00</td>\n","      <td>64.00</td>\n","      <td>47.117104</td>\n","      <td>47.117104</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>8.309008e+07</td>\n","      <td>9.5</td>\n","      <td>10.392305</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>314309</th>\n","      <td>0.015934</td>\n","      <td>62.28</td>\n","      <td>6.00</td>\n","      <td>65.72</td>\n","      <td>31.289158</td>\n","      <td>31.289158</td>\n","      <td>0.0</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>1.0</td>\n","      <td>...</td>\n","      <td>3.514623</td>\n","      <td>54.59</td>\n","      <td>8.334500e+07</td>\n","      <td>9.5</td>\n","      <td>10.532031</td>\n","      <td>4.974893</td>\n","      <td>69.757079</td>\n","      <td>0.19</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>66880</th>\n","      <td>0.000000</td>\n","      <td>53.46</td>\n","      <td>5.94</td>\n","      <td>63.36</td>\n","      <td>0.602080</td>\n","      <td>0.602080</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.249154</td>\n","      <td>54.06</td>\n","      <td>8.333107e+07</td>\n","      <td>9.5</td>\n","      <td>10.399888</td>\n","      <td>0.352811</td>\n","      <td>0.426318</td>\n","      <td>0.15</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>160088</th>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>6.00</td>\n","      <td>64.00</td>\n","      <td>3.005428</td>\n","      <td>3.005428</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>1.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>8.309440e+07</td>\n","      <td>9.5</td>\n","      <td>10.392305</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>94890</th>\n","      <td>0.064326</td>\n","      <td>167.40</td>\n","      <td>16.84</td>\n","      <td>64.00</td>\n","      <td>2.415185</td>\n","      <td>2.415185</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>3.037121</td>\n","      <td>161.24</td>\n","      <td>8.300769e+07</td>\n","      <td>9.5</td>\n","      <td>17.965543</td>\n","      <td>4.303642</td>\n","      <td>84.921741</td>\n","      <td>0.11</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>284143</th>\n","      <td>0.012008</td>\n","      <td>57.16</td>\n","      <td>6.00</td>\n","      <td>64.00</td>\n","      <td>9.014776</td>\n","      <td>9.014776</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.330928</td>\n","      <td>54.20</td>\n","      <td>8.307202e+07</td>\n","      <td>9.5</td>\n","      <td>10.400347</td>\n","      <td>0.469517</td>\n","      <td>1.578792</td>\n","      <td>0.07</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>140498</th>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>6.00</td>\n","      <td>64.00</td>\n","      <td>43.054271</td>\n","      <td>43.054271</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>54.00</td>\n","      <td>8.295084e+07</td>\n","      <td>9.5</td>\n","      <td>10.392305</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","    <tr>\n","      <th>61862</th>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>1.00</td>\n","      <td>64.00</td>\n","      <td>3.764777</td>\n","      <td>3.764777</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>0.0</td>\n","      <td>...</td>\n","      <td>0.000000</td>\n","      <td>42.00</td>\n","      <td>8.312811e+07</td>\n","      <td>9.5</td>\n","      <td>9.165151</td>\n","      <td>0.000000</td>\n","      <td>0.000000</td>\n","      <td>0.00</td>\n","      <td>141.55</td>\n","      <td>Attack</td>\n","    </tr>\n","  </tbody>\n","</table>\n","<p>10 rows × 47 columns</p>\n","</div>"],"text/plain":["        flow_duration  Header_Length  Protocol Type  Duration         Rate  \\\n","413784       0.130570       30076.00          17.00     64.00  4580.297292   \n","433656       0.000000           0.00           1.00     64.00    37.205150   \n","148341       0.000000          54.00           6.00     64.00    47.117104   \n","314309       0.015934          62.28           6.00     65.72    31.289158   \n","66880        0.000000          53.46           5.94     63.36     0.602080   \n","160088       0.000000          54.00           6.00     64.00     3.005428   \n","94890        0.064326         167.40          16.84     64.00     2.415185   \n","284143       0.012008          57.16           6.00     64.00     9.014776   \n","140498       0.000000          54.00           6.00     64.00    43.054271   \n","61862        0.000000           0.00           1.00     64.00     3.764777   \n","\n","              Srate  Drate  fin_flag_number  syn_flag_number  rst_flag_number  \\\n","413784  4580.297292    0.0              0.0              0.0              0.0   \n","433656    37.205150    0.0              0.0              0.0              0.0   \n","148341    47.117104    0.0              0.0              1.0              0.0   \n","314309    31.289158    0.0              1.0              0.0              1.0   \n","66880      0.602080    0.0              0.0              0.0              0.0   \n","160088     3.005428    0.0              0.0              1.0              0.0   \n","94890      2.415185    0.0              0.0              0.0              0.0   \n","284143     9.014776    0.0              0.0              0.0              0.0   \n","140498    43.054271    0.0              0.0              0.0              0.0   \n","61862      3.764777    0.0              0.0              0.0              0.0   \n","\n","        ...       Std  Tot size           IAT  Number   Magnitue    Radius  \\\n","413784  ...  0.000000     50.00  8.310187e+07     9.5  10.000000  0.000000   \n","433656  ...  0.000000     42.00  8.314968e+07     9.5   9.165151  0.000000   \n","148341  ...  0.000000     54.00  8.309008e+07     9.5  10.392305  0.000000   \n","314309  ...  3.514623     54.59  8.334500e+07     9.5  10.532031  4.974893   \n","66880   ...  0.249154     54.06  8.333107e+07     9.5  10.399888  0.352811   \n","160088  ...  0.000000     54.00  8.309440e+07     9.5  10.392305  0.000000   \n","94890   ...  3.037121    161.24  8.300769e+07     9.5  17.965543  4.303642   \n","284143  ...  0.330928     54.20  8.307202e+07     9.5  10.400347  0.469517   \n","140498  ...  0.000000     54.00  8.295084e+07     9.5  10.392305  0.000000   \n","61862   ...  0.000000     42.00  8.312811e+07     9.5   9.165151  0.000000   \n","\n","        Covariance  Variance  Weight   label  \n","413784    0.000000      0.00  141.55  Attack  \n","433656    0.000000      0.00  141.55  Attack  \n","148341    0.000000      0.00  141.55  Attack  \n","314309   69.757079      0.19  141.55  Attack  \n","66880     0.426318      0.15  141.55  Attack  \n","160088    0.000000      0.00  141.55  Attack  \n","94890    84.921741      0.11  141.55  Attack  \n","284143    1.578792      0.07  141.55  Attack  \n","140498    0.000000      0.00  141.55  Attack  \n","61862     0.000000      0.00  141.55  Attack  \n","\n","[10 rows x 47 columns]"]},"execution_count":6,"metadata":{},"output_type":"execute_result"}],"source":["df.head(10)"]},{"cell_type":"code","execution_count":7,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":6,"status":"ok","timestamp":1705010666108,"user":{"displayName":"Hiếu Võ","userId":"16914104909337499796"},"user_tz":-420},"id":"s7vQhiYs2Qr9","outputId":"6aeaa4e2-0e19-41a5-9647-3c349a3f4d18"},"outputs":[{"name":"stdout","output_type":"stream","text":["[[0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]\n"," [0 1]]\n","413784    Attack\n","433656    Attack\n","148341    Attack\n","314309    Attack\n","66880     Attack\n","160088    Attack\n","94890     Attack\n","284143    Attack\n","140498    Attack\n","61862     Attack\n","Name: label, dtype: object\n"]},{"data":{"text/plain":["array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1])"]},"execution_count":7,"metadata":{},"output_type":"execute_result"}],"source":["y = pd.get_dummies(df['label']).iloc[:, ::-1].to_numpy()\n","\n","y_labels = np.argmax(y, axis=1)\n","\n","print(y[0:10])\n","print(df['label'].head(10))\n","y_labels[:10]"]},{"cell_type":"code","execution_count":8,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["Label 0: 10800 occurrences\n","Label 1: 446692 occurrences\n"]}],"source":["\n","# Đếm số lượng mỗi nhãn\n","label_counts = np.bincount(y_labels)\n","\n","# In ra số lượng mỗi nhãn\n","for i, count in enumerate(label_counts):\n","    print(f\"Label {i}: {count} occurrences\")"]},{"cell_type":"code","execution_count":9,"metadata":{},"outputs":[],"source":["df.replace([np.inf, -np.inf], np.nan, inplace=True)\n","df.fillna(method='ffill', inplace=True)\n","scaler = MinMaxScaler()\n","X = df.iloc[:, :-1]\n","X = scaler.fit_transform(X)"]},{"cell_type":"code","execution_count":10,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":10,"status":"ok","timestamp":1705010285692,"user":{"displayName":"Hiếu Võ","userId":"16914104909337499796"},"user_tz":-420},"id":"pE1Z1nAX3YYS","outputId":"4e882cad-be9f-4132-82c8-e4337fabbbd3"},"outputs":[{"data":{"text/plain":["array([[1.90806842e-06, 3.06411609e-03, 3.61702128e-01, 2.50980392e-01,\n","        6.24015984e-04, 6.24015984e-04, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 7.81040287e-03, 2.37079184e-03, 2.64137986e-04,\n","        1.02313499e-03, 0.00000000e+00, 6.12745098e-04, 4.95717949e-01,\n","        6.29629630e-01, 7.46665227e-03, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01],\n","       [0.00000000e+00, 0.00000000e+00, 2.12765957e-02, 2.50980392e-01,\n","        5.06879945e-06, 5.06879945e-06, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 6.45207194e-03, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 4.96003158e-01,\n","        6.29629630e-01, 1.38777878e-17, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01],\n","       [0.00000000e+00, 5.50147190e-06, 1.27659574e-01, 2.50980392e-01,\n","        6.41919596e-06, 6.41919596e-06, 0.00000000e+00, 0.00000000e+00,\n","        1.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.25000000e-01,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.48956834e-03, 3.55618777e-03, 3.96206979e-04,\n","        1.53470249e-03, 0.00000000e+00, 9.19117647e-04, 4.95647618e-01,\n","        6.29629630e-01, 1.09753170e-02, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01],\n","       [2.32846470e-07, 6.34503092e-06, 1.27659574e-01, 2.57725490e-01,\n","        4.26280950e-06, 4.26280950e-06, 0.00000000e+00, 1.00000000e+00,\n","        0.00000000e+00, 1.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 2.47500000e-01, 0.00000000e+00,\n","        1.27577320e-02, 1.90240654e-05, 8.80523912e-06, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.67084084e-03, 3.55618777e-03, 7.66330331e-04,\n","        1.73071806e-03, 3.19619791e-04, 9.64307598e-04, 4.97168238e-01,\n","        6.29629630e-01, 1.22249845e-02, 3.19906947e-04, 5.08120993e-07,\n","        1.90000000e-01, 5.76970443e-01],\n","       [0.00000000e+00, 5.44645718e-06, 1.26382979e-01, 2.48470588e-01,\n","        8.20268566e-08, 8.20268566e-08, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 3.13897080e-04, 1.08964834e-04, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.50412189e-03, 3.55618777e-03, 4.25922502e-04,\n","        1.54478860e-03, 2.26580851e-05, 9.23713235e-04, 4.97085158e-01,\n","        6.29629630e-01, 1.10431350e-02, 2.26872874e-05, 3.10536150e-09,\n","        1.50000000e-01, 5.76970443e-01],\n","       [0.00000000e+00, 5.50147190e-06, 1.27659574e-01, 2.50980392e-01,\n","        4.09457052e-07, 4.09457052e-07, 0.00000000e+00, 0.00000000e+00,\n","        1.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.25000000e-01,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.48956834e-03, 3.55618777e-03, 3.96206979e-04,\n","        1.53470249e-03, 0.00000000e+00, 9.19117647e-04, 4.95673382e-01,\n","        6.29629630e-01, 1.09753170e-02, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01],\n","       [9.40014050e-07, 1.70545629e-05, 3.58297872e-01, 2.50980392e-01,\n","        3.29042781e-07, 3.29042781e-07, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 2.66873219e-02, 3.21716453e-02, 4.05451808e-03,\n","        1.52687210e-02, 2.76195769e-04, 9.13296569e-03, 4.95156112e-01,\n","        6.29629630e-01, 7.87082408e-02, 2.76742686e-04, 6.18582662e-07,\n","        1.10000000e-01, 5.76970443e-01],\n","       [1.75482470e-07, 5.82340988e-06, 1.27659574e-01, 2.50980392e-01,\n","        1.22816581e-06, 1.22816581e-06, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 6.25000000e-03,\n","        0.00000000e+00, 0.00000000e+00, 3.30196467e-06, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.51220719e-03, 3.55618777e-03, 4.42431126e-04,\n","        1.54538417e-03, 3.00946302e-05, 9.34436275e-04, 4.95539844e-01,\n","        6.29629630e-01, 1.10472466e-02, 3.01919711e-05, 1.15001556e-08,\n","        7.00000000e-02, 5.76970443e-01],\n","       [0.00000000e+00, 5.50147190e-06, 1.27659574e-01, 2.50980392e-01,\n","        5.86567889e-06, 5.86567889e-06, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 8.48956834e-03, 3.55618777e-03, 3.96206979e-04,\n","        1.53470249e-03, 0.00000000e+00, 9.19117647e-04, 4.94817024e-01,\n","        6.29629630e-01, 1.09753170e-02, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01],\n","       [0.00000000e+00, 0.00000000e+00, 2.12765957e-02, 2.50980392e-01,\n","        5.12910141e-07, 5.12910141e-07, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 1.00000000e+00, 1.00000000e+00,\n","        1.00000000e+00, 6.45207194e-03, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 4.95874444e-01,\n","        6.29629630e-01, 1.38777878e-17, 0.00000000e+00, 0.00000000e+00,\n","        0.00000000e+00, 5.76970443e-01]])"]},"execution_count":10,"metadata":{},"output_type":"execute_result"}],"source":["X[0:10]"]},{"cell_type":"code","execution_count":11,"metadata":{},"outputs":[],"source":["X_train, X_test, y_train, y_test = train_test_split(X, y_labels, test_size=0.2, random_state=75)"]},{"cell_type":"code","execution_count":12,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["Label 0: 2153 occurrences\n","Label 1: 89346 occurrences\n"]}],"source":["\n","# Đếm số lượng mỗi nhãn\n","label_counts = np.bincount(y_test)\n","#label_counts = np.bincount(y_train)\n","\n","# In ra số lượng mỗi nhãn\n","for i, count in enumerate(label_counts):\n","    print(f\"Label {i}: {count} occurrences\")"]},{"cell_type":"code","execution_count":13,"metadata":{"colab":{"background_save":true},"id":"2MbooI7p5BKT"},"outputs":[],"source":["from sklearn.metrics import classification_report, confusion_matrix\n","\n","def eval(model):\n","    # Dự đoán nhãn trên tập kiểm thử\n","    y_pred = model.predict(X_test)\n","    y_pred = (y_pred > 0.5).astype(int)\n","    y_pred = tf.squeeze(y_pred, axis=1)\n","    print(y_pred[:10])\n","    print(y_test[:10])\n","    cm = confusion_matrix(y_test, y_pred)\n","    # Tính toán độ chính xác tổng thể\n","    total_accuracy = (cm[0, 0] + cm[1, 1]) / np.sum(cm)\n","\n","\n","    # Tính toán F1-score, Recall và Precision\n","    report = classification_report(y_test, y_pred, target_names=None)\n","\n","    # Thêm vào báo cáo\n","    report += '\\nTotal accuracy: {:0.2f} \\n'.format(total_accuracy)\n","\n","    print(report)\n"]},{"cell_type":"code","execution_count":14,"metadata":{},"outputs":[],"source":["def build_model():\n","  n_features = X_train.shape[1]\n","  model = Sequential()\n","  model.add(Dense(512, input_dim=n_features, activation='relu'))  # Input layer và hidden layer đầu tiên\n","  model.add(Dropout(0.3))  # Thêm dropout để giảm overfitting\n","  model.add(Dense(256, activation='relu'))  # Thêm hidden layer thứ hai\n","  model.add(Dropout(0.3))  # Thêm dropout\n","  model.add(Dense(128, activation='relu'))  # Thêm hidden layer thứ hai\n","  model.add(Dropout(0.3))  # Thêm dropout\n","  model.add(Dense(64, activation='relu'))  # Thêm hidden layer thứ hai\n","  model.add(Dropout(0.3))  # Thêm dropout\n","  model.add(Dense(1, activation='sigmoid'))  # Output layer\n","\n","  # Compile mô hình\n","  model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])\n","\n","  return model"]},{"cell_type":"code","execution_count":15,"metadata":{},"outputs":[{"data":{"text/plain":["'\\nfrom tensorflow.keras.models import Sequential\\nfrom tensorflow.keras.layers import Dense, Dropout\\n\\nmodel = build_model()\\n\\n# Huấn luyện mô hình\\nmodel.fit(X_train, y_train, epochs=10, batch_size=128, validation_data=(X_test, y_test))\\n\\neval(model)\\n'"]},"execution_count":15,"metadata":{},"output_type":"execute_result"}],"source":["'''\n","from tensorflow.keras.models import Sequential\n","from tensorflow.keras.layers import Dense, Dropout\n","\n","model = build_model()\n","\n","# Huấn luyện mô hình\n","model.fit(X_train, y_train, epochs=10, batch_size=128, validation_data=(X_test, y_test))\n","\n","eval(model)\n","'''"]},{"cell_type":"code","execution_count":16,"metadata":{},"outputs":[{"data":{"text/plain":["'\\neval(model)\\n'"]},"execution_count":16,"metadata":{},"output_type":"execute_result"}],"source":["'''\n","eval(model)\n","'''"]},{"cell_type":"code","execution_count":17,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["365993\n"]}],"source":["# Tính tổng số mẫu\n","\n","n_samples = X_train.shape[0]\n","print(n_samples)\n","\n","# Tính chỉ số cho việc chia tách\n","# Phần 1: 0 -> 5/10 * n_samples\n","# Phần 2: 5/10 * n_samples -> (5+3)/10 * n_samples\n","# Phần 3: (5+3)/10 * n_samples -> n_samples\n","indices = [int(5/10 * n_samples), int((5+3)/10 * n_samples)]\n","\n","# Chia dữ liệu\n","client_data = np.split(X_train, indices)\n","client_labels = np.split(y_train, indices)\n","\n","# client_data[0], client_labels[0] sẽ chứa 50% dữ liệu đầu tiên\n","# client_data[1], client_labels[1] sẽ chứa 30% dữ liệu tiếp theo\n","# client_data[2], client_labels[2] sẽ chứa 20% dữ liệu cuối cùng\n"]},{"cell_type":"code","execution_count":18,"metadata":{},"outputs":[{"data":{"text/plain":["'\\nimport numpy as np\\nn_samples = X_train.shape[0]\\n# Giả sử X_train, y_train là dữ liệu huấn luyện của bạn\\n# n_samples = X_train.shape[0]\\n\\n# Định nghĩa tỷ lệ cho mỗi node\\nratios = [0.3, 0.3, 0.4]  # Tổng cộng phải bằng 1\\n\\n# Tính chỉ số cho việc chia tách dựa trên tỷ lệ\\nindices = [int(sum(ratios[:i+1]) * n_samples) for i in range(len(ratios)-1)]\\n\\n# Chia dữ liệu\\nclient_data = np.split(X_train, indices)\\nclient_labels = np.split(y_train, indices)\\n\\n# In kết quả để kiểm tra\\nfor i, (data, labels) in enumerate(zip(client_data, client_labels)):\\n    print(f\"Node {i+1}:\")\\n    print(f\" - Dữ liệu: {data.shape}\")\\n    print(f\" - Nhãn: {labels.shape}\")\\n\\n'"]},"execution_count":18,"metadata":{},"output_type":"execute_result"}],"source":["'''\n","import numpy as np\n","n_samples = X_train.shape[0]\n","# Giả sử X_train, y_train là dữ liệu huấn luyện của bạn\n","# n_samples = X_train.shape[0]\n","\n","# Định nghĩa tỷ lệ cho mỗi node\n","ratios = [0.3, 0.3, 0.4]  # Tổng cộng phải bằng 1\n","\n","# Tính chỉ số cho việc chia tách dựa trên tỷ lệ\n","indices = [int(sum(ratios[:i+1]) * n_samples) for i in range(len(ratios)-1)]\n","\n","# Chia dữ liệu\n","client_data = np.split(X_train, indices)\n","client_labels = np.split(y_train, indices)\n","\n","# In kết quả để kiểm tra\n","for i, (data, labels) in enumerate(zip(client_data, client_labels)):\n","    print(f\"Node {i+1}:\")\n","    print(f\" - Dữ liệu: {data.shape}\")\n","    print(f\" - Nhãn: {labels.shape}\")\n","\n","'''"]},{"cell_type":"code","execution_count":19,"metadata":{},"outputs":[],"source":["def fed_avg(models, data_distribution):\n","    \"\"\"\n","    models: danh sách các mô hình từ các node\n","    data_distribution: danh sách chứa số lượng dữ liệu mỗi node có\n","    \"\"\"\n","    # Tính tổng số mẫu dữ liệu\n","    total_data = sum(data_distribution)\n","\n","    # Khởi tạo trọng số trung bình\n","    avg_weights = [np.zeros_like(w) for w in models[0].get_weights()]\n","\n","    # Cộng trọng số từ tất cả mô hình, đã được nhân với tỷ lệ dữ liệu tương ứng\n","    for model, data_count in zip(models, data_distribution):\n","        for idx, layer_weights in enumerate(model.get_weights()):\n","            avg_weights[idx] += (layer_weights * data_count) / total_data\n","\n","    return avg_weights\n"]},{"cell_type":"code","execution_count":20,"metadata":{},"outputs":[],"source":["global_model = build_model()"]},{"cell_type":"code","execution_count":21,"metadata":{},"outputs":[],"source":["# Các thông số DP\n","l2_norm_clip = 1.0\n","noise_multiplier = 0.4\n","\n","\n","# FL\n","num_rounds = 50\n","num_epochs = 3\n","batch_size = 128"]},{"cell_type":"code","execution_count":22,"metadata":{},"outputs":[{"data":{"text/plain":["'from art.estimators.classification import TensorFlowV2Classifier\\nfrom art.attacks.evasion import FastGradientMethod\\nimport numpy as np\\n\\n# Xác định mô hình và định nghĩa bộ phân loại ART\\nmodel = build_model()  # Sử dụng hàm build_model bạn đã cung cấp\\nclassifier = TensorFlowV2Classifier(model=model, nb_classes=2, input_shape=(X_train.shape[1],), loss_object=tf.keras.losses.BinaryCrossentropy(), clip_values=(0, 1))\\n\\n# Tạo một instance của FGSM\\nattack = FastGradientMethod(estimator=classifier, eps=0.1)\\n\\n# Chọn một phần của tập dữ liệu để tạo dữ liệu đầu độc\\n# Sử dụng một phần nhỏ của dữ liệu để tránh tác động quá lớn đến quá trình huấn luyện\\nindices = np.random.choice(len(X_train), size=100000, replace=False)\\nX_train_subset = X_train[indices]\\ny_train_subset = y_train[indices]\\n\\n# Tạo dữ liệu đầu độc\\npoisoned_X = attack.generate(x=X_train_subset)\\npoisoned_y = y_train_subset  # Giữ nguyên nhãn của các mẫu đã chọn\\n\\n# Bây giờ `poisoned_X` và `poisoned_y` là dữ liệu và nhãn đầu độc\\n'"]},"execution_count":22,"metadata":{},"output_type":"execute_result"}],"source":["\"\"\"from art.estimators.classification import TensorFlowV2Classifier\n","from art.attacks.evasion import FastGradientMethod\n","import numpy as np\n","\n","# Xác định mô hình và định nghĩa bộ phân loại ART\n","model = build_model()  # Sử dụng hàm build_model bạn đã cung cấp\n","classifier = TensorFlowV2Classifier(model=model, nb_classes=2, input_shape=(X_train.shape[1],), loss_object=tf.keras.losses.BinaryCrossentropy(), clip_values=(0, 1))\n","\n","# Tạo một instance của FGSM\n","attack = FastGradientMethod(estimator=classifier, eps=0.1)\n","\n","# Chọn một phần của tập dữ liệu để tạo dữ liệu đầu độc\n","# Sử dụng một phần nhỏ của dữ liệu để tránh tác động quá lớn đến quá trình huấn luyện\n","indices = np.random.choice(len(X_train), size=100000, replace=False)\n","X_train_subset = X_train[indices]\n","y_train_subset = y_train[indices]\n","\n","# Tạo dữ liệu đầu độc\n","poisoned_X = attack.generate(x=X_train_subset)\n","poisoned_y = y_train_subset  # Giữ nguyên nhãn của các mẫu đã chọn\n","\n","# Bây giờ `poisoned_X` và `poisoned_y` là dữ liệu và nhãn đầu độc\n","\"\"\""]},{"cell_type":"code","execution_count":23,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["109797\n","109797\n"]}],"source":["import numpy as np\n","\n","# Giả sử `X_train` và `y_train` là dữ liệu và nhãn huấn luyện ban đầu\n","\n","# Chọn một phần nhỏ của tập dữ liệu để thay đổi nhãn\n","num_poisoned_samples = int(0.3 * len(X_train))  # Ví dụ: 10% dữ liệu\n","poisoned_indices = np.random.choice(len(X_train), num_poisoned_samples, replace=False)\n","\n","# Sao chép dữ liệu và nhãn ban đầu để tránh thay đổi trực tiếp\n","poisoned_X = np.copy(X_train[poisoned_indices])\n","poisoned_y = np.copy(y_train[poisoned_indices])\n","\n","# Thay đổi nhãn của dữ liệu đã chọn\n","# Ví dụ: đối với bài toán phân loại nhị phân, chúng ta có thể đảo ngược nhãn\n","poisoned_y = 1 - poisoned_y\n","\n","# Bây giờ `poisoned_X` và `poisoned_y` là dữ liệu và nhãn đã bị đầu độc\n","print(len(poisoned_X))\n","print(len(poisoned_y))"]},{"cell_type":"code","execution_count":24,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["Round  0  - Client  1\n","Round  0  - Client  2\n","Round  0  - Client  3\n","Round  0  - Evaluation:\n","2860/2860 [==============================] - 6s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.65      0.99      0.79      2153\n","           1       1.00      0.99      0.99     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.83      0.99      0.89     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n","Round  1  - Client  1\n","Round  1  - Client  2\n","Round  1  - Client  3\n","Round  1  - Evaluation:\n","2860/2860 [==============================] - 6s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.71      0.95      0.81      2153\n","           1       1.00      0.99      0.99     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.86      0.97      0.90     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n","Round  2  - Client  1\n","Round  2  - Client  2\n","Round  2  - Client  3\n","Round  2  - Evaluation:\n","2860/2860 [==============================] - 6s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.69      0.98      0.81      2153\n","           1       1.00      0.99      0.99     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.84      0.99      0.90     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n"]}],"source":["for round in range(3):\n","    local_models = []\n","    cnt = 1\n","    for client_X, client_y in zip(client_data, client_labels):\n","        print('Round ', round, ' - Client ', cnt)\n","        cnt += 1\n","        # Sao chép trọng số từ global model\n","        local_model = build_model()\n","        local_model.set_weights(global_model.get_weights())\n","        \n","        # Biên dịch mô hình với bộ tối ưu hóa DP\n","        local_model.compile(optimizer=Adam(), loss='binary_crossentropy', metrics=['accuracy'])\n","\n","\n","        # Huấn luyện mô hình cục bộ\n","        local_model.fit(client_X, client_y, epochs=num_epochs, batch_size=batch_size, verbose=0)\n","        local_models.append(local_model)\n","\n","    # Cập nhật global model dựa trên trung bình trọng số từ các mô hình cục bộ\n","    global_model.set_weights(fed_avg(local_models, [len(cli_dt) for cli_dt in client_data]))\n","\n","    print('Round ', round, ' - Evaluation:')\n","    eval(global_model)"]},{"cell_type":"code","execution_count":25,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["Round  0  - Evaluation:\n","[0.9257112, 0.9448498, 0.9533578, 0.92191905]\n","Mean:  0.9364595\n","Eliminated: Node  0\n","Eliminated: Node  3\n","2860/2860 [==============================] - 5s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.76      0.90      0.83      2153\n","           1       1.00      0.99      1.00     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.88      0.95      0.91     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n","Round  1  - Evaluation:\n","[0.93602586, 0.9495543, 0.96481246, 0.9353792]\n","Mean:  0.94644296\n","Eliminated: Node  0\n","Eliminated: Node  3\n","2860/2860 [==============================] - 5s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.75      0.92      0.83      2153\n","           1       1.00      0.99      1.00     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.88      0.96      0.91     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n","Round  2  - Evaluation:\n","[0.94758564, 0.9647006, 0.97294444, 0.9563727]\n","Mean:  0.9604009\n","Eliminated: Node  0\n","Eliminated: Node  3\n","2860/2860 [==============================] - 6s 2ms/step\n","tf.Tensor([1 1 1 1 1 1 1 1 1 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 1 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.72      0.97      0.82      2153\n","           1       1.00      0.99      0.99     89346\n","\n","    accuracy                           0.99     91499\n","   macro avg       0.86      0.98      0.91     91499\n","weighted avg       0.99      0.99      0.99     91499\n","\n","Total accuracy: 0.99 \n","\n"]}],"source":["# Bước 2: Thêm dữ liệu đầu độc vào node độc hại\n","# Giả sử 'poisoned_X' và 'poisoned_y' là dữ liệu và nhãn đầu độc\n","client_data.append(poisoned_X)  # Thêm dữ liệu đầu độc\n","client_labels.append(poisoned_y)  # Thêm nhãn đầu độc\n","\n","def cosine_similarity(weights_a, weights_b):\n","    dot_product = np.sum([np.dot(a.flatten(), b.flatten()) for a, b in zip(weights_a, weights_b)])\n","    norm_a = np.sqrt(np.sum([np.dot(a.flatten(), a.flatten()) for a in weights_a]))\n","    norm_b = np.sqrt(np.sum([np.dot(b.flatten(), b.flatten()) for b in weights_b]))\n","    return dot_product / (norm_a * norm_b)\n","\n","# Giả định global_weights là trọng số của mô hình toàn cầu từ vòng lặp trước\n","global_weights_previous_round = global_model.get_weights()\n","\n","for round in range(num_rounds):\n","    local_models = []\n","    similarities = []\n","    for client_X, client_y in zip(client_data, client_labels):\n","        # Huấn luyện mô hình cục bộ như bình thường\n","        local_model = build_model()\n","        local_model.set_weights(global_weights_previous_round)\n","        local_model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])\n","        local_model.fit(client_X, client_y, epochs=num_epochs, batch_size=batch_size, verbose=0)\n","        \n","        # Tính cosine similarity giữa trọng số mô hình cục bộ và mô hình toàn cầu\n","        local_weights = local_model.get_weights()\n","        sim = cosine_similarity(local_weights, global_weights_previous_round)\n","        similarities.append(sim)\n","        local_models.append(local_model)\n","\n","    # Tính giá trị trung bình của các điểm cosine similarity\n","    mean_similarity = np.mean(similarities)\n","\n","    # Lọc và giữ lại những mô hình có similarity cao hơn giá trị trung bình\n","    valid_models = [local_models[i] for i, sim in enumerate(similarities) if sim >= mean_similarity]\n","    data_distribution = [len(client_data[i]) for i, sim in enumerate(similarities) if sim >= mean_similarity]\n","    # Cập nhật trọng số mô hình toàn cầu chỉ từ những mô hình hợp lệ\n","    print('Round ', round, ' - Evaluation:')\n","    print(similarities)\n","    print('Mean: ', mean_similarity)\n","    for i, sim in enumerate(similarities):\n","        if sim < mean_similarity:\n","            print('Eliminated: Node ', i)\n","    if valid_models:\n","        global_model.set_weights(fed_avg(valid_models, data_distribution))\n","        eval(global_model)\n","    else:\n","        print('ERR')\n","    # Lưu lại trọng số mới cho vòng lặp tiếp theo\n","    global_weights_previous_round = global_model.get_weights()\n"]},{"cell_type":"code","execution_count":null,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["Round  0  - Client  1\n","Round  0  - Client  2\n","Round  0  - Client  3\n","Round  0  - Evaluation:\n","141597/141597 [==============================] - 214s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.70      0.57      0.63    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.84      0.78      0.81   4531078\n","weighted avg       0.98      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  1  - Client  1\n","Round  1  - Client  2\n","Round  1  - Client  3\n","Round  1  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.59      0.82      0.69    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.79      0.91      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  2  - Client  1\n","Round  2  - Client  2\n","Round  2  - Client  3\n","Round  2  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.58      0.84      0.69    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.79      0.91      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  3  - Client  1\n","Round  3  - Client  2\n","Round  3  - Client  3\n","Round  3  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.59      0.84      0.69    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.79      0.91      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  4  - Client  1\n","Round  4  - Client  2\n","Round  4  - Client  3\n","Round  4  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.57      0.88      0.69    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.78      0.93      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  5  - Client  1\n","Round  5  - Client  2\n","Round  5  - Client  3\n","Round  5  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.58      0.87      0.69    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.79      0.93      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  6  - Client  1\n","Round  6  - Client  2\n","Round  6  - Client  3\n","Round  6  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.57      0.91      0.70    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.78      0.94      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  7  - Client  1\n","Round  7  - Client  2\n","Round  7  - Client  3\n","Round  7  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.55      0.92      0.69    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.78      0.95      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  8  - Client  1\n","Round  8  - Client  2\n","Round  8  - Client  3\n","Round  8  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.56      0.92      0.70    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.78      0.95      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  9  - Client  1\n","Round  9  - Client  2\n","Round  9  - Client  3\n","Round  9  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.81      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.90      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  10  - Client  1\n","Round  10  - Client  2\n","Round  10  - Client  3\n","Round  10  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.85      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  11  - Client  1\n","Round  11  - Client  2\n","Round  11  - Client  3\n","Round  11  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.62      0.85      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  12  - Client  1\n","Round  12  - Client  2\n","Round  12  - Client  3\n","Round  12  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.75      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.87      0.85   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  13  - Client  1\n","Round  13  - Client  2\n","Round  13  - Client  3\n","Round  13  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.67      0.70      0.68    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.83      0.85      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  14  - Client  1\n","Round  14  - Client  2\n","Round  14  - Client  3\n","Round  14  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.81      0.73    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.90      0.86   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  15  - Client  1\n","Round  15  - Client  2\n","Round  15  - Client  3\n","Round  15  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.78      0.71    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.88      0.85   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  16  - Client  1\n","Round  16  - Client  2\n","Round  16  - Client  3\n","Round  16  - Evaluation:\n","141597/141597 [==============================] - 214s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.62      0.81      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.90      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  17  - Client  1\n","Round  17  - Client  2\n","Round  17  - Client  3\n","Round  17  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.80      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.90      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  18  - Client  1\n","Round  18  - Client  2\n","Round  18  - Client  3\n","Round  18  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.60      0.87      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.93      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  19  - Client  1\n","Round  19  - Client  2\n","Round  19  - Client  3\n","Round  19  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.77      0.71    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.88      0.85   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  20  - Client  1\n","Round  20  - Client  2\n","Round  20  - Client  3\n","Round  20  - Evaluation:\n","141597/141597 [==============================] - 219s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.85      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.92      0.86   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  21  - Client  1\n","Round  21  - Client  2\n","Round  21  - Client  3\n","Round  21  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.81      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.90      0.86   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  22  - Client  1\n","Round  22  - Client  2\n","Round  22  - Client  3\n","Round  22  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.82      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.82      0.91      0.86   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  23  - Client  1\n","Round  23  - Client  2\n","Round  23  - Client  3\n","Round  23  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.65      0.82      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.82      0.90      0.86   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  24  - Client  1\n","Round  24  - Client  2\n","Round  24  - Client  3\n","Round  24  - Evaluation:\n","141597/141597 [==============================] - 221s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.62      0.89      0.73    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.94      0.86   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  25  - Client  1\n","Round  25  - Client  2\n","Round  25  - Client  3\n","Round  25  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.80      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.89      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  26  - Client  1\n","Round  26  - Client  2\n","Round  26  - Client  3\n","Round  26  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.65      0.78      0.71    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.82      0.88      0.85   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  27  - Client  1\n","Round  27  - Client  2\n","Round  27  - Client  3\n","Round  27  - Evaluation:\n","141597/141597 [==============================] - 219s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.84      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  28  - Client  1\n","Round  28  - Client  2\n","Round  28  - Client  3\n","Round  28  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.87      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.93      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  29  - Client  1\n","Round  29  - Client  2\n","Round  29  - Client  3\n","Round  29  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.62      0.85      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  30  - Client  1\n","Round  30  - Client  2\n","Round  30  - Client  3\n","Round  30  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.84      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.91      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  31  - Client  1\n","Round  31  - Client  2\n","Round  31  - Client  3\n","Round  31  - Evaluation:\n","141597/141597 [==============================] - 214s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.60      0.87      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.93      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  32  - Client  1\n","Round  32  - Client  2\n","Round  32  - Client  3\n","Round  32  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.79      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.89      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  33  - Client  1\n","Round  33  - Client  2\n","Round  33  - Client  3\n","Round  33  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.62      0.84      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  34  - Client  1\n","Round  34  - Client  2\n","Round  34  - Client  3\n","Round  34  - Evaluation:\n","141597/141597 [==============================] - 219s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.78      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.88      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  35  - Client  1\n","Round  35  - Client  2\n","Round  35  - Client  3\n","Round  35  - Evaluation:\n","141597/141597 [==============================] - 218s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.66      0.77      0.71    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.88      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  36  - Client  1\n","Round  36  - Client  2\n","Round  36  - Client  3\n","Round  36  - Evaluation:\n","141597/141597 [==============================] - 220s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.81      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.90      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  37  - Client  1\n","Round  37  - Client  2\n","Round  37  - Client  3\n","Round  37  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.74      0.68    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.86      0.83   4531078\n","weighted avg       0.98      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  38  - Client  1\n","Round  38  - Client  2\n","Round  38  - Client  3\n","Round  38  - Evaluation:\n","141597/141597 [==============================] - 221s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.60      0.91      0.72    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.95      0.86   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  39  - Client  1\n","Round  39  - Client  2\n","Round  39  - Client  3\n","Round  39  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.65      0.78      0.71    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.88      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  40  - Client  1\n","Round  40  - Client  2\n","Round  40  - Client  3\n","Round  40  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.63      0.79      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.81      0.89      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  41  - Client  1\n","Round  41  - Client  2\n","Round  41  - Client  3\n","Round  41  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.83      0.70    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.91      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  42  - Client  1\n","Round  42  - Client  2\n","Round  42  - Client  3\n","Round  42  - Evaluation:\n","141597/141597 [==============================] - 217s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.78      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.88      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  43  - Client  1\n","Round  43  - Client  2\n","Round  43  - Client  3\n","Round  43  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.64      0.81      0.71    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.82      0.90      0.85   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  44  - Client  1\n","Round  44  - Client  2\n","Round  44  - Client  3\n","Round  44  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.61      0.83      0.70    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.91      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  45  - Client  1\n","Round  45  - Client  2\n","Round  45  - Client  3\n","Round  45  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.60      0.85      0.70    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.92      0.85   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  46  - Client  1\n","Round  46  - Client  2\n","Round  46  - Client  3\n","Round  46  - Evaluation:\n","141597/141597 [==============================] - 223s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.68      0.70      0.69    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.83      0.85      0.84   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  47  - Client  1\n","Round  47  - Client  2\n","Round  47  - Client  3\n","Round  47  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.68      0.72      0.70    106786\n","           1       0.99      0.99      0.99   4424292\n","\n","    accuracy                           0.99   4531078\n","   macro avg       0.84      0.85      0.84   4531078\n","weighted avg       0.99      0.99      0.99   4531078\n","\n","Total accuracy: 0.99 \n","\n","Round  48  - Client  1\n","Round  48  - Client  2\n","Round  48  - Client  3\n","Round  48  - Evaluation:\n","141597/141597 [==============================] - 215s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.60      0.83      0.70    106786\n","           1       1.00      0.99      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.80      0.91      0.84   4531078\n","weighted avg       0.99      0.98      0.98   4531078\n","\n","Total accuracy: 0.98 \n","\n","Round  49  - Client  1\n","Round  49  - Client  2\n","Round  49  - Client  3\n","Round  49  - Evaluation:\n","141597/141597 [==============================] - 216s 2ms/step\n","tf.Tensor([1 0 1 1 1 1 1 1 0 1], shape=(10,), dtype=int64)\n","[1 1 1 1 1 1 1 1 0 1]\n","              precision    recall  f1-score   support\n","\n","           0       0.59      0.97      0.73    106786\n","           1       1.00      0.98      0.99   4424292\n","\n","    accuracy                           0.98   4531078\n","   macro avg       0.79      0.98      0.86   4531078\n","weighted avg       0.99      0.98      0.99   4531078\n","\n","Total accuracy: 0.98 \n","\n"]}],"source":["\"\"\"\n","global_model = build_model()\n","\n","for round in range(num_rounds):\n","    local_models = []\n","    cnt = 1\n","    for client_X, client_y in zip(client_data, client_labels):\n","        print('Round ', round, ' - Client ', cnt)\n","        # Tính số batch đầy đủ có thể có\n","        num_full_batches = client_X.shape[0] // batch_size\n","        # Cắt dữ liệu để chỉ bao gồm các batch đầy đủ\n","        full_batch_data_size = num_full_batches * batch_size\n","        client_X_full_batches = client_X[:full_batch_data_size]\n","        client_y_full_batches = client_y[:full_batch_data_size]\n","\n","        cnt += 1\n","        local_model = build_model()\n","        local_model.set_weights(global_model.get_weights())\n","        \n","        optimizer = DPKerasSGDOptimizer(\n","            l2_norm_clip=l2_norm_clip,\n","            noise_multiplier=noise_multiplier,\n","            num_microbatches=1,\n","            learning_rate=0.001\n","        )\n","\n","        local_model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['accuracy'])\n","\n","        local_model.fit(client_X_full_batches, client_y_full_batches, epochs=num_epochs, batch_size=batch_size, verbose=0)\n","        local_models.append(local_model)\n","\n","    global_model.set_weights(fed_avg(local_models, [len(cli_dt[:full_batch_data_size]) for cli_dt in client_data]))\n","\n","    print('Round ', round, ' - Evaluation:')\n","    eval(global_model)\n","\"\"\""]}],"metadata":{"colab":{"authorship_tag":"ABX9TyMSDjvZ+MB3aEVf9ABRJwOt","mount_file_id":"1kw32fVpav4IPJJtK8yC8brhEcbQqd3Tn","provenance":[]},"kernelspec":{"display_name":"Python 3","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":"3.9.16"}},"nbformat":4,"nbformat_minor":0}