malicious_url_detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of FORENZIKA_SEMINAR.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ed_ilACFfJ_Y",
        "colab_type": "code",
        "outputId": "faaa0beb-26ae-4f58-df68-83ee0b2022e0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "import torch\n",
        "\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
        "n_gpu = torch.cuda.device_count()\n",
        "torch.cuda.get_device_name(0)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Tesla P4'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rWmRPX3ENRhy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import pandas as pd \n",
        "\n",
        "data = pd.read_csv(\"kaggle.csv\") "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jhx81_yTNSc0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "benign=data[data['label']=='good']\n",
        "malicious=data[data['label']=='bad']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4rjbyVSb7pY7",
        "colab_type": "code",
        "outputId": "db613601-213f-4f2c-cdfa-9a6d2440b75c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "malicious.count()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "url      75643\n",
              "label    75643\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SG8XYY-N8WRQ",
        "colab_type": "code",
        "outputId": "896a5d29-aeda-4004-f841-1538b31ce8e9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "benign.count()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "url      344821\n",
              "label    344821\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T55zA8JR7RhG",
        "colab_type": "code",
        "outputId": "13954053-bdd3-419c-873d-5e09195c6386",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "ratio=len(malicious)/len(benign)\n",
        "print(ratio)\n",
        "benign=benign.sample(frac=ratio)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.21936888994579795\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kHA9QCqO7cBm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X=benign.append(malicious)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dezaPxrJ9B81",
        "colab_type": "code",
        "outputId": "000ce896-4015-4e2f-fc10-883c39df855d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "X.count()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "url      420464\n",
              "label    420464\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u2oW0ies9Voy",
        "colab_type": "code",
        "outputId": "7d6bb785-d6cf-457e-ccc7-c7ada5a1d6dc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "X.iloc[40000]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "url      en.wikipedia.org/wiki/Nesty\n",
              "label                           good\n",
              "Name: 82767, dtype: object"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IBOaQhze9rqz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "vocab={c:i+2 for i,c in enumerate('''0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ-._~:/?#[]@!$&'()*+,;%=''')}\n",
        "vocab['[PAD]']=0\n",
        "vocab['[UNK]']=1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "amg2lDJ7-yAY",
        "colab_type": "code",
        "outputId": "eb56bd26-34af-40b4-ba58-f61871097c70",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "print(vocab)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'0': 2, '1': 3, '2': 4, '3': 5, '4': 6, '5': 7, '6': 8, '7': 9, '8': 10, '9': 11, 'a': 12, 'b': 13, 'c': 14, 'd': 15, 'e': 16, 'f': 17, 'g': 18, 'h': 19, 'i': 20, 'j': 21, 'k': 22, 'l': 23, 'm': 24, 'n': 25, 'o': 26, 'p': 27, 'q': 28, 'r': 29, 's': 30, 't': 31, 'u': 32, 'v': 33, 'w': 34, 'x': 35, 'y': 36, 'z': 37, 'A': 38, 'B': 39, 'C': 40, 'D': 41, 'E': 42, 'F': 43, 'G': 44, 'H': 45, 'I': 46, 'J': 47, 'K': 48, 'L': 49, 'M': 50, 'N': 51, 'O': 52, 'P': 53, 'Q': 54, 'R': 55, 'S': 56, 'T': 57, 'U': 58, 'V': 59, 'W': 60, 'X': 61, 'Y': 62, 'Z': 63, '-': 64, '.': 65, '_': 66, '~': 67, ':': 68, '/': 69, '?': 70, '#': 71, '[': 72, ']': 73, '@': 74, '!': 75, '$': 76, '&': 77, \"'\": 78, '(': 79, ')': 80, '*': 81, '+': 82, ',': 83, ';': 84, '%': 85, '=': 86, '[PAD]': 0, '[UNK]': 1}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y4zhC8miXvuO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class CharVocab():\n",
        "  def __init__(self,vocab):\n",
        "    self.vocab=vocab\n",
        "  def ctoi(self,c):\n",
        "    try:\n",
        "      return self.vocab[c]\n",
        "    except:\n",
        "      return self.vocab['[UNK]']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ADRZl__wX_60",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "char_vocab=CharVocab(vocab)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sJD0etcOOUWf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "SEQUENCE_LENGTH=128\n",
        "\n",
        "examples=[]\n",
        "for index,row in X.iterrows():\n",
        "  tokens=[char_vocab.ctoi(c) for c in row['url']][:SEQUENCE_LENGTH]\n",
        "  tokens+=[char_vocab.ctoi('[PAD]') for i in range(SEQUENCE_LENGTH-len(tokens))]\n",
        "  examples.append((tokens,int(row['label']=='bad')))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mgknUDv5ZALr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torch.nn.functional as F\n",
        "\n",
        "class URLGledac(torch.nn.Module):\n",
        "  def __init__(self,sequence_length=128,embedding_size=32,hidden_size=128):\n",
        "    super(URLGledac,self).__init__()\n",
        "\n",
        "    self.embedding=torch.nn.Embedding(len(vocab),embedding_size)\n",
        "    self.lstm=torch.nn.LSTM(embedding_size,hidden_size,bidirectional=True,num_layers=2,dropout=0.2)\n",
        "    self.pool=torch.nn.MaxPool1d(sequence_length)\n",
        "    self.dense=torch.nn.Linear(hidden_size*2,128)\n",
        "    self.batch=torch.nn.BatchNorm1d(128)\n",
        "    self.dropout=torch.nn.Dropout(0.2)\n",
        "    self.output=torch.nn.Linear(128,2)\n",
        "\n",
        "  def forward(self,url):\n",
        "    #print(url.shape)\n",
        "    embedding=self.embedding(url)\n",
        "    #print(embedding.shape)\n",
        "    out,_=self.lstm(torch.transpose(embedding,0,1))\n",
        "    #print(out.shape)\n",
        "    out=self.pool(torch.transpose(out,0,2))\n",
        "    out=self.dense(torch.transpose(out.squeeze(dim=2),0,1))\n",
        "    out=F.relu(out)\n",
        "    out=self.batch(out)\n",
        "    out=self.output(out)\n",
        "\n",
        "    scores=torch.nn.functional.log_softmax(out,dim=1)\n",
        "\n",
        "    return scores\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rzHLTsI2buEM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.metrics import precision_recall_fscore_support\n",
        "\n",
        "def accuracy(predictions,labels):\n",
        "  return accuracy_score(labels,predictions)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hVMma5OzKRiU",
        "colab_type": "code",
        "outputId": "f992c97a-6ed3-466e-882a-1768e8da0072",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        }
      },
      "source": [
        "import numpy as np\n",
        "import random\n",
        "\n",
        "folds=5\n",
        "random.shuffle(examples)\n",
        "ratio=int(1/folds*len(examples))\n",
        "X_folds=[]\n",
        "for split in range(folds):\n",
        "  train_set=examples[:split*ratio]+examples[(split+1)*ratio:]\n",
        "  test_set=examples[split*ratio:(split+1)*ratio]\n",
        "  X_folds.append((train_set,test_set))\n",
        "  print(len(train_set))\n",
        "  print(len(test_set))\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "336372\n",
            "84092\n",
            "336372\n",
            "84092\n",
            "336372\n",
            "84092\n",
            "336372\n",
            "84092\n",
            "336372\n",
            "84092\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "X4IChGBBagzi",
        "colab_type": "code",
        "outputId": "bd557a2f-d11e-4ee8-cfef-4cd7778b1c4e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "import random\n",
        "\n",
        "url_gledac=URLGledac()\n",
        "url_gledac.cuda()\n",
        "\n",
        "url_gledac.train()\n",
        "loss=torch.nn.NLLLoss()\n",
        "optimizer=torch.optim.Adam(url_gledac.parameters(),lr=5e-5,betas=(0.8,0.999))\n",
        "\n",
        "BATCH_SIZE=128\n",
        "for train_set, test_set in X_folds[0:1]:\n",
        "\n",
        "  best_val_acc=0.\n",
        "  for epoch in range(10):\n",
        "    train_accuracy=0.\n",
        "    train_loss=0.\n",
        "    counter=0\n",
        "\n",
        "    random.seed(69)\n",
        "    random.shuffle(train_set)\n",
        "\n",
        "    url_gledac.train()\n",
        "    num_batches=len(train_set)//BATCH_SIZE\n",
        "    for it in range(num_batches):\n",
        "    #for url,label in train_set:\n",
        "      urls=[url[0] for url in train_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "      labels=[url[1] for url in train_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "\n",
        "      counter+=1\n",
        "      if counter%500==0:\n",
        "        print('Train accuracy: %s loss: %s'%(train_accuracy/counter,train_loss/counter))\n",
        "        counter=0\n",
        "        train_accuracy=0.\n",
        "        train_loss=0.\n",
        "\n",
        "      #url=torch.tensor(url).unsqueeze(0).to(device)\n",
        "      urls=torch.tensor(urls).to(device)\n",
        "      scores=url_gledac(urls)\n",
        "\n",
        "      #train_accuracy+=1. if torch.max(scores.cpu(),1)[1].item()==label else 0.\n",
        "      #print(scores.shape)\n",
        "      preds=torch.max(scores.cpu(),1)[1].detach().numpy()\n",
        "      train_accuracy+=accuracy(labels,preds)\n",
        "      \n",
        "      labels=torch.tensor(labels).to(device)\n",
        "      #label=torch.tensor(label).unsqueeze(0).to(device)\n",
        "      loss_val=loss(scores,labels)\n",
        "\n",
        "      train_loss+=loss_val.item()\n",
        "\n",
        "      #print(loss_val)\n",
        "      loss_val.backward()\n",
        "      optimizer.step()\n",
        "\n",
        "      \n",
        "    print('Validation...')\n",
        "\n",
        "    url_gledac.eval()\n",
        "    validation_accuracy=0.\n",
        "    num_batches=len(test_set)//BATCH_SIZE\n",
        "    for it in range(num_batches):\n",
        "    #for url,label in train_set:\n",
        "      urls=[url[0] for url in test_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "      labels=[url[1] for url in test_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "      with torch.no_grad():\n",
        "        urls=torch.tensor(urls).to(device)\n",
        "        scores=url_gledac(urls)\n",
        "        preds=torch.max(scores.cpu(),1)[1].detach().numpy()\n",
        "        validation_accuracy+=accuracy(labels,preds)\n",
        "\n",
        "    print('Validation accuracy: %s'%(validation_accuracy/num_batches)) \n",
        "    if best_val_acc<validation_accuracy/num_batches:\n",
        "      best_val_acc=validation_accuracy/num_batches\n",
        "      torch.save(url_gledac.state_dict(), 'forenzika_model'+str(round(validation_accuracy/num_batches,5)))\n",
        "  print()\n",
        "  print(\"Best result for fold: %s\"%(best_val_acc))\n",
        "  print()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train accuracy: 0.813203125 loss: 0.4392803165912628\n",
            "Train accuracy: 0.93453125 loss: 0.19806267829239368\n",
            "Train accuracy: 0.95365625 loss: 0.12928789917379618\n",
            "Train accuracy: 0.96071875 loss: 0.11096702927350997\n",
            "Train accuracy: 0.964578125 loss: 0.10460640705935657\n",
            "Validation...\n",
            "Validation accuracy: 0.9683093559451219\n",
            "Train accuracy: 0.967171875 loss: 0.0953153676148504\n",
            "Train accuracy: 0.970734375 loss: 0.09284277200233192\n",
            "Train accuracy: 0.974328125 loss: 0.08274199733883142\n",
            "Train accuracy: 0.97490625 loss: 0.08330525983963162\n",
            "Train accuracy: 0.975421875 loss: 0.0745852038115263\n",
            "Validation...\n",
            "Validation accuracy: 0.979027724847561\n",
            "Train accuracy: 0.9773125 loss: 0.06094888239819556\n",
            "Train accuracy: 0.980046875 loss: 0.05818241689167917\n",
            "Train accuracy: 0.9810625 loss: 0.055900570208206775\n",
            "Train accuracy: 0.98178125 loss: 0.05409956355020404\n",
            "Train accuracy: 0.981546875 loss: 0.054647663303650916\n",
            "Validation...\n",
            "Validation accuracy: 0.9825528772865854\n",
            "Train accuracy: 0.982578125 loss: 0.04660733804386109\n",
            "Train accuracy: 0.985359375 loss: 0.043793804452754555\n",
            "Train accuracy: 0.984328125 loss: 0.047136506888084115\n",
            "Train accuracy: 0.98446875 loss: 0.046189133367501196\n",
            "Train accuracy: 0.984265625 loss: 0.047312693124637005\n",
            "Validation...\n",
            "Validation accuracy: 0.9836128048780488\n",
            "Train accuracy: 0.984671875 loss: 0.03985739319212735\n",
            "Train accuracy: 0.987703125 loss: 0.03756695321388542\n",
            "Train accuracy: 0.986765625 loss: 0.040814667286351324\n",
            "Train accuracy: 0.984859375 loss: 0.04367175699025393\n",
            "Train accuracy: 0.984515625 loss: 0.04569597680401057\n",
            "Validation...\n",
            "Validation accuracy: 0.9816358612804879\n",
            "Train accuracy: 0.984890625 loss: 0.03909657986834645\n",
            "Train accuracy: 0.98640625 loss: 0.04025945595465601\n",
            "Train accuracy: 0.986359375 loss: 0.040705302267335354\n",
            "Train accuracy: 0.984875 loss: 0.04366591374669224\n",
            "Train accuracy: 0.98540625 loss: 0.0417352458210662\n",
            "Validation...\n",
            "Validation accuracy: 0.9814572217987805\n",
            "Train accuracy: 0.983578125 loss: 0.042829076381400225\n",
            "Train accuracy: 0.98478125 loss: 0.04476323949638754\n",
            "Train accuracy: 0.983921875 loss: 0.04761797110643238\n",
            "Train accuracy: 0.9835625 loss: 0.047785027137957514\n",
            "Train accuracy: 0.9826875 loss: 0.0506390972500667\n",
            "Validation...\n",
            "Validation accuracy: 0.9793492759146342\n",
            "Train accuracy: 0.979875 loss: 0.05341107946820557\n",
            "Train accuracy: 0.981046875 loss: 0.05487309439852834\n",
            "Train accuracy: 0.978953125 loss: 0.06241656509228051\n",
            "Train accuracy: 0.977609375 loss: 0.06506495585478843\n",
            "Train accuracy: 0.9758125 loss: 0.07055901399627328\n",
            "Validation...\n",
            "Validation accuracy: 0.9733708079268293\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-116-c3989dacd4da>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     40\u001b[0m       \u001b[0;31m#train_accuracy+=1. if torch.max(scores.cpu(),1)[1].item()==label else 0.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     41\u001b[0m       \u001b[0;31m#print(scores.shape)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 42\u001b[0;31m       \u001b[0mpreds\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mscores\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcpu\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdetach\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnumpy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     43\u001b[0m       \u001b[0mtrain_accuracy\u001b[0m\u001b[0;34m+=\u001b[0m\u001b[0maccuracy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlabels\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mpreds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     44\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ayu1Ve9BjJ4O",
        "colab_type": "text"
      },
      "source": [
        "##T-SNE"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Vx_f-nhy-Lzt",
        "colab_type": "code",
        "outputId": "32ce3962-3a28-46f7-8e66-a84f0f61c84a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 170
        }
      },
      "source": [
        "model_path='forenzika_model0.98361'\n",
        "\n",
        "url_gledac=URLGledac()\n",
        "url_gledac.cuda()\n",
        "url_gledac.load_state_dict(torch.load(model_path))\n",
        "url_gledac.eval()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "URLGledac(\n",
              "  (embedding): Embedding(87, 32)\n",
              "  (lstm): LSTM(32, 128, num_layers=2, dropout=0.2, bidirectional=True)\n",
              "  (pool): MaxPool1d(kernel_size=128, stride=128, padding=0, dilation=1, ceil_mode=False)\n",
              "  (dense): Linear(in_features=256, out_features=128, bias=True)\n",
              "  (batch): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "  (dropout): Dropout(p=0.2, inplace=False)\n",
              "  (output): Linear(in_features=128, out_features=2, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 117
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9XjgUVEZiRnE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "from sklearn.manifold import TSNE\n",
        "from sklearn.decomposition import PCA\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "chars=[char_vocab.ctoi(c) for c in '''0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ-._~:/?#[]@!$&'()*+,;%=''']\n",
        "labels=[c for c in '''0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ-._~:/?#[]@!$&'()*+,;%=''']\n",
        "embedding=url_gledac.embedding(torch.tensor(chars).to(device))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UeS-HNBejcal",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "embedding=embedding.cpu().detach().numpy()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5VUpo3szjock",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "tsne=TSNE(perplexity=5,learning_rate=10.,random_state=69)\n",
        "new_embeddings=tsne.fit_transform(embedding)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oroZGDptj_aL",
        "colab_type": "code",
        "outputId": "a0cfd8db-26b9-4539-9e0e-8990c604b1aa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        }
      },
      "source": [
        "plt.figure(figsize=(8,8))\n",
        "plt.scatter(new_embeddings[:,0],new_embeddings[:,1],marker='.',s=0.01)\n",
        "\n",
        "for i,e in enumerate(new_embeddings):\n",
        "  plt.annotate(labels[i],xy=e)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAegAAAHSCAYAAAAnsVjHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIBW0iIiIB6mgRUREPEgFLSIi4kH/H5bcaRVQCXyTAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "38mZ79_Aa3uU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "from sklearn import metrics\n",
        "\n",
        "url_gledac.eval()\n",
        "pred=[]\n",
        "y=[]\n",
        "num_batches=len(test_set)//BATCH_SIZE\n",
        "for it in range(num_batches):\n",
        "  urls=[url[0] for url in test_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "  labels=[url[1] for url in test_set[it*BATCH_SIZE:(it+1)*BATCH_SIZE]]\n",
        "  with torch.no_grad():\n",
        "    urls=torch.tensor(urls).to(device)\n",
        "    scores=url_gledac(urls)\n",
        "    preds=torch.max(scores.cpu(),1)[1].detach()\n",
        "    pred.append(preds)\n",
        "    labels=torch.tensor(labels)\n",
        "    y.append(labels)\n",
        "pred=torch.cat(pred,0).squeeze().numpy()\n",
        "y=torch.cat(y,0).squeeze().numpy()\n",
        "\n",
        "fpr, tpr, thresholds = metrics.roc_curve(y, pred, pos_label=1)\n",
        "auc=metrics.auc(fpr, tpr)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "m1g5uQJuZwEU",
        "colab_type": "code",
        "outputId": "373f6f36-840c-4b9e-c374-2d80650aa5ea",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 501
        }
      },
      "source": [
        "plt.figure(figsize=(10,8))\n",
        "plt.plot(fpr, tpr, 'b', label = 'AUC = %0.2f' % auc)\n",
        "plt.legend(loc = 'lower right')\n",
        "plt.plot([0, 1], [0, 1],'r--')\n",
        "plt.xlim([0, 1])\n",
        "plt.ylim([0, 1])\n",
        "plt.ylabel('True Positive Rate')\n",
        "plt.xlabel('False Positive Rate')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmkAAAHkCAYAAABlt7hJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMV3lDwqffwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 720x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z1PY227XDkGX",
        "colab_type": "code",
        "outputId": "19267b5d-3a51-4dad-a9bc-9bf56a367023",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        }
      },
      "source": [
        "precision_recall_fscore_support(y,pred)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(array([0.98713715, 0.96689693]),\n",
              " array([0.9929793 , 0.94064808]),\n",
              " array([0.99004961, 0.95359191]),\n",
              " array([68939, 15029]))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 120
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "58l0P3MZB66D",
        "colab_type": "code",
        "outputId": "d340c185-8435-4c0f-ae56-7ec65cf30f81",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "precision_recall_fscore_support(y,pred,average='macro')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.9770170399623457, 0.966813690459497, 0.971820756450223, None)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 122
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VHdpfn2uI4Ae",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}