gdpr_tlm_blog_post/gdpr_tlm_blog_post.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from cleanlab_tlm import TLM\n",
    "import pandas as pd\n",
    "from typing import List, Tuple"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set your API key\n",
    "import os\n",
    "os.environ[\"CLEANLAB_TLM_API_KEY\"] = \"<API key>\"  # Get your API key from: https://tlm.cleanlab.ai/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load log entries dataset\n",
    "df = pd.read_csv(\"application_logs.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Querying TLM... 100%|██████████|"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "GDPR Violation: True\n",
      "\n",
      "The log entry contains personal data, specifically the full name and date of birth of a new user. Under the General Data Protection Regulation (GDPR), personal data must be processed lawfully, fairly, and transparently. Logging such information without a valid legal basis, appropriate security measures, or without the user's consent could be considered a violation of GDPR principles. Additionally, personal data should be minimized and only collected if necessary for the intended purpose.\n",
      "\n",
      "Trustworthiness Score: 0.8558611725369876\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Initialize TLM with high quality preset and GPT-4o\n",
    "tlm = TLM(quality_preset=\"high\", options={\"model\": \"gpt-4o\", \"log\": [\"explanation\"]})\n",
    "\n",
    "log_entry = \"2024-08-24 19:32:18 WARN: New user registered: Charlie Smith, DOB: 1981-03-26\"\n",
    "prompt = [f\"Analyze this log entry for GDPR violations: {log_entry}\\nViolates GDPR (True/False):\"]\n",
    "\n",
    "result = tlm.try_prompt(prompt)\n",
    "print(f\"\\nGDPR Violation: {result[0]['response']}\\n\")\n",
    "print(f\"Trustworthiness Score: {result[0]['trustworthiness_score']}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_gdpr_violation(response: str) -> bool:\n",
    "    \"\"\"\n",
    "    Extracts the GDPR violation status from the response.\n",
    "    \n",
    "    Args:\n",
    "        response (str): Response from the TLM\n",
    "    \n",
    "    Returns:\n",
    "        bool: True if violates GDPR, False otherwise\n",
    "    \"\"\"\n",
    "    return response.strip().lower() == 'true'\n",
    "\n",
    "def classify_gdpr_violations(texts: List[str], prompt_template: str) -> Tuple[List[bool], List[float], List[str]]:\n",
    "    \"\"\"\n",
    "    Classifies a list of log entries for GDPR violations using TLM.\n",
    "    \n",
    "    Args:\n",
    "        texts (List[str]): List of log entries to classify\n",
    "        prompt_template (str): Template string for formatting the prompt\n",
    "    \n",
    "    Returns:\n",
    "        Tuple[List[bool], List[float], List[str]]: A tuple containing three lists:\n",
    "            - List of GDPR violation statuses for each log entry\n",
    "            - List of trustworthiness scores for each prediction\n",
    "            - List of low trustworthiness explanations (empty string if trustworthiness is high)\n",
    "    \"\"\"\n",
    "    prompts = [prompt_template.format(text=text) for text in texts]\n",
    "    outputs = tlm.try_prompt(prompts)\n",
    "    \n",
    "    violations = [parse_gdpr_violation(output['response']) for output in outputs]\n",
    "    trustworthiness_scores = [output['trustworthiness_score'] for output in outputs]\n",
    "    low_trust_explanations = [output['log']['explanation'] if output['trustworthiness_score'] < 0.5 else '' for output in outputs]\n",
    "    \n",
    "    return violations, trustworthiness_scores, low_trust_explanations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the prompt template\n",
    "prompt_template = \"\"\"\n",
    "Analyze the following application log entry and determine if it violates GDPR by containing or processing personal data without proper safeguards.\n",
    "Consider things like user identifiers, names, email addresses, IP addresses, or any other personally identifiable information.\n",
    "Respond with only 'True' if it violates GDPR, or 'False' if it complies with GDPR.\n",
    "\n",
    "Log entry: {text}\n",
    "\n",
    "Violates GDPR (True/False):\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Querying TLM... 100%|██████████|"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 49.9 s, sys: 12.8 s, total: 1min 2s\n",
      "Wall time: 2min 20s\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# Classify all log entries in batch\n",
    "violations, trustworthiness_scores, low_trust_explanations = classify_gdpr_violations(df[\"log_entry\"].tolist(), prompt_template)\n",
    "\n",
    "# Add results to the DataFrame\n",
    "df['violates_GDPR'] = violations\n",
    "df['trustworthiness'] = trustworthiness_scores\n",
    "df['low_trust_explanation'] = low_trust_explanations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Log Entries with High Trustworthiness Scores:\n",
      "================================================================================\n",
      "\n",
      "Log Entry:\n",
      "2024-08-17 17:48:26 INFO: User 'user250@example.com' changed password\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9998444409832958\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-05 02:16:10 ERROR: New user registered: Bob Brown, DOB: 1976-01-02\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9998299213052612\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-28 15:28:47 INFO: New user registered: Jane Brown, DOB: 1985-11-21\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9998298630687377\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-17 01:10:32 DEBUG: Geo-location data updated for user 'user_140': lat\n",
      "-71.9892, long 162.0127\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.999812477946249\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-06 01:00:18 DEBUG: Password reset requested for account:\n",
      "user155@example.com\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997828234259813\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-13 10:08:10 DEBUG: User 'user374@example.com' exported customer data\n",
      "report\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997755336983166\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-21 03:47:43 INFO: User 'user178@example.com' exported customer data\n",
      "report\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997709243712074\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-12 00:33:20 WARN: New user registered: Charlie Brown, DOB: 1962-06-24\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997582632395947\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-31 09:18:41 WARN: User 'Bob Johnson' (ID: U12661) granted admin\n",
      "privileges\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997418946084226\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-22 10:21:44 DEBUG: User 'Alice Johnson' (ID: U1845) granted admin\n",
      "privileges\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9997325926289524\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "import textwrap\n",
    "\n",
    "# Filter for examples with high trustworthiness scores\n",
    "high_trust_threshold = 0.8\n",
    "high_trust_logs = df[df['trustworthiness'] >= high_trust_threshold].sort_values('trustworthiness', ascending=False)\n",
    "\n",
    "print(\"\\nLog Entries with High Trustworthiness Scores:\")\n",
    "print(\"=\" * 80)\n",
    "\n",
    "for i, (_, row) in enumerate(high_trust_logs.iterrows()):\n",
    "    if i >= 10:\n",
    "        break\n",
    "    print(\"\\nLog Entry:\")\n",
    "    print(textwrap.fill(row['log_entry'], width=80))\n",
    "    print(f\"\\nViolates GDPR: {row['violates_GDPR']}\")\n",
    "    print(f\"Trustworthiness score: {row['trustworthiness']}\")\n",
    "    print(\"-\" * 80)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Top GDPR-violating log entries (high trustworthiness):\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>log_entry</th>\n",
       "      <th>violates_GDPR</th>\n",
       "      <th>trustworthiness</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>2024-09-14 22:23:46 DEBUG: Password reset requ...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.998532</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>2024-09-11 16:27:24 INFO: Geo-location data up...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.995860</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>2024-08-30 10:43:05 DEBUG: User login successf...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.993578</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>2024-08-24 01:31:02 WARN: Newsletter subscript...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.999325</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>2024-08-24 19:32:18 WARN: New user registered:...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.996044</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            log_entry  violates_GDPR  \\\n",
       "9   2024-09-14 22:23:46 DEBUG: Password reset requ...           True   \n",
       "15  2024-09-11 16:27:24 INFO: Geo-location data up...           True   \n",
       "22  2024-08-30 10:43:05 DEBUG: User login successf...           True   \n",
       "23  2024-08-24 01:31:02 WARN: Newsletter subscript...           True   \n",
       "28  2024-08-24 19:32:18 WARN: New user registered:...           True   \n",
       "\n",
       "    trustworthiness  \n",
       "9          0.998532  \n",
       "15         0.995860  \n",
       "22         0.993578  \n",
       "23         0.999325  \n",
       "28         0.996044  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "# Filter for GDPR violations with high trustworthiness scores\n",
    "gdpr_violations = df[(df['violates_GDPR'] == True) & (df['trustworthiness'] >= high_trust_threshold)]\n",
    "\n",
    "print(\"\\nTop GDPR-violating log entries (high trustworthiness):\")\n",
    "gdpr_violations[['log_entry', 'violates_GDPR', 'trustworthiness']].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Log Entries with Low Trustworthiness Scores:\n",
      "================================================================================\n",
      "\n",
      "Log Entry:\n",
      "2024-09-09 12:25:00 INFO: Unauthorized access attempt to admin panel from IP:\n",
      "53.125.88.241\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.2336285708561434\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. However, this assessment is\n",
      "not entirely accurate. The log entry contains an IP address, which can be\n",
      "considered personal data under GDPR if it can be linked to an individual.\n",
      "However, the mere presence of an IP address in a log entry does not\n",
      "automatically constitute a GDPR violation. The context of how the IP address is\n",
      "used, stored, and protected is crucial. If the IP address is used solely for\n",
      "security purposes and is adequately protected, it may not violate GDPR.\n",
      "Therefore, the answer \"True\" is not definitively correct without additional\n",
      "context.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-30 14:48:51 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "8.118.70.251\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.3933484367633626\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer provided is \"True,\" indicating that the log entry violates GDPR.\n",
      "However, this conclusion is not necessarily accurate. The log entry contains an\n",
      "IP address, which can be considered personal data under GDPR if it can be linked\n",
      "to an individual. However, the mere presence of an IP address in a log entry\n",
      "does not automatically constitute a GDPR violation. The context of how the IP\n",
      "address is used, stored, and protected is crucial. If proper safeguards and\n",
      "justifications for processing the IP address are in place, it may comply with\n",
      "GDPR. Therefore, the answer is not fully accurate and lacks the necessary\n",
      "context to make a definitive judgment.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-26 06:07:39 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "74.62.73.229\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.40525719781303643\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. However, this assessment is\n",
      "not entirely accurate. The log entry contains an IP address, which can be\n",
      "considered personal data under GDPR if it can be linked to an individual.\n",
      "However, the mere presence of an IP address in a log entry does not\n",
      "automatically constitute a GDPR violation. The context of how the data is used,\n",
      "stored, and protected is crucial. If the IP address is used solely for security\n",
      "purposes and is adequately protected, it may not violate GDPR. Therefore, the\n",
      "answer \"True\" is not definitively correct without additional context.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-22 00:01:31 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "156.41.24.195\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.40636479363904343\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer provided is \"True,\" indicating that the log entry violates GDPR.\n",
      "However, this conclusion is not necessarily accurate. The log entry contains an\n",
      "IP address, which can be considered personal data under GDPR if it can be linked\n",
      "to an individual. However, the mere presence of an IP address in a log entry\n",
      "does not automatically constitute a GDPR violation. The context of how the data\n",
      "is processed, stored, and protected also matters. The log entry does not provide\n",
      "enough information to definitively conclude that GDPR is violated. Therefore,\n",
      "the answer is not fully accurate or justified.  This response is untrustworthy\n",
      "due to lack of consistency in possible responses from the model. Here's one\n",
      "inconsistent alternate response that the model considered (which may not be\n",
      "accurate either):  False.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-06 00:29:17 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "78.29.194.147\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.4127898836101229\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. The log entry includes an IP\n",
      "address (78.29.194.147), which is considered personal data under GDPR as it can\n",
      "be used to identify an individual. However, the answer does not provide any\n",
      "reasoning or context to support the conclusion, making it difficult to assess\n",
      "its accuracy and completeness. Additionally, the log entry itself does not\n",
      "provide enough information to definitively determine whether proper safeguards\n",
      "are in place or not. Therefore, while the answer addresses the user's question,\n",
      "it lacks the necessary explanation and confidence to be considered fully\n",
      "accurate and helpful.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-09-15 06:52:55 INFO: Unusual activity detected from IP: 133.202.221.230\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.49032043086015886\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. However, this assessment is\n",
      "not entirely accurate. The log entry contains an IP address, which can be\n",
      "considered personal data under GDPR if it can be linked to an individual.\n",
      "However, the mere presence of an IP address in a log entry does not\n",
      "automatically constitute a GDPR violation. The context of how the IP address is\n",
      "used, stored, and protected, as well as whether proper safeguards and legal\n",
      "bases for processing are in place, are crucial factors in determining\n",
      "compliance. Therefore, the answer \"True\" is overly simplistic and does not fully\n",
      "consider these nuances.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-22 15:20:50 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "190.248.208.39\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.4910890893010253\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer provided is \"True,\" indicating that the log entry violates GDPR.\n",
      "However, this conclusion is not necessarily accurate. The log entry contains an\n",
      "IP address, which can be considered personal data under GDPR if it can be linked\n",
      "to an individual. However, the mere presence of an IP address in a log entry\n",
      "does not automatically constitute a GDPR violation. The context of how the data\n",
      "is processed, stored, and protected is crucial. If proper safeguards and\n",
      "justifications for logging the IP address are in place, it may comply with GDPR.\n",
      "Therefore, the answer is not definitively nd lacks the necessary context to make\n",
      "a conclusive determination.\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Log Entry:\n",
      "2024-08-17 16:13:32 ERROR: Unauthorized access attempt to admin panel from IP:\n",
      "117.59.245.203\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.49124775309238206\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. However, this assessment is\n",
      "not entirely accurate. The log entry contains an IP address (117.59.245.203),\n",
      "which can be considered personal data under GDPR if it can be linked to an\n",
      "individual. However, the mere presence of an IP address in a log entry does not\n",
      "automatically constitute a GDPR violation. The context of how the IP address is\n",
      "used, stored, and protected is crucial. If the IP address is used solely for\n",
      "security purposes and is adequately protected, it may not violate GDPR.\n",
      "Therefore, the answer \"True\" is not definitively correct without additional\n",
      "context.  This response is untrustworthy due to lack of consistency in possible\n",
      "responses from the model. Here's one inconsistent alternate response that the\n",
      "model considered (which may not be accurate either):  False.\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "import textwrap\n",
    "\n",
    "# Filter for examples with low trustworthiness scores\n",
    "low_trust_threshold = 0.5\n",
    "low_trust_logs = df[(df['violates_GDPR'] == True) & \n",
    "                    (df['trustworthiness'] < low_trust_threshold)].sort_values('trustworthiness')\n",
    "\n",
    "print(\"\\nLog Entries with Low Trustworthiness Scores:\")\n",
    "print(\"=\" * 80)\n",
    "\n",
    "for _, row in low_trust_logs.iterrows():\n",
    "    print(\"\\nLog Entry:\")\n",
    "    print(textwrap.fill(row['log_entry'], width=80))\n",
    "    print(f\"\\nViolates GDPR: {row['violates_GDPR']}\")\n",
    "    print(f\"Trustworthiness score: {row['trustworthiness']}\")\n",
    "    print(\"\\nLow trustworthiness explanation:\")\n",
    "    print(textwrap.fill(row['low_trust_explanation'], width=80))\n",
    "    print(\"-\" * 80)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Top low-trustworthiness log entries:\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>log_entry</th>\n",
       "      <th>violates_GDPR</th>\n",
       "      <th>trustworthiness</th>\n",
       "      <th>low_trust_explanation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>432</th>\n",
       "      <td>2024-09-09 12:25:00 INFO: Unauthorized access ...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.233629</td>\n",
       "      <td>The answer \"True\" indicates that the log entry...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>542</th>\n",
       "      <td>2024-08-30 14:48:51 ERROR: Unauthorized access...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.393348</td>\n",
       "      <td>The answer provided is \"True,\" indicating that...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>891</th>\n",
       "      <td>2024-08-26 06:07:39 ERROR: Unauthorized access...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.405257</td>\n",
       "      <td>The answer \"True\" indicates that the log entry...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>954</th>\n",
       "      <td>2024-08-22 00:01:31 ERROR: Unauthorized access...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.406365</td>\n",
       "      <td>The answer provided is \"True,\" indicating that...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>800</th>\n",
       "      <td>2024-09-06 00:29:17 ERROR: Unauthorized access...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.412790</td>\n",
       "      <td>The answer \"True\" indicates that the log entry...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             log_entry  violates_GDPR  \\\n",
       "432  2024-09-09 12:25:00 INFO: Unauthorized access ...           True   \n",
       "542  2024-08-30 14:48:51 ERROR: Unauthorized access...           True   \n",
       "891  2024-08-26 06:07:39 ERROR: Unauthorized access...           True   \n",
       "954  2024-08-22 00:01:31 ERROR: Unauthorized access...           True   \n",
       "800  2024-09-06 00:29:17 ERROR: Unauthorized access...           True   \n",
       "\n",
       "     trustworthiness                              low_trust_explanation  \n",
       "432         0.233629  The answer \"True\" indicates that the log entry...  \n",
       "542         0.393348  The answer provided is \"True,\" indicating that...  \n",
       "891         0.405257  The answer \"True\" indicates that the log entry...  \n",
       "954         0.406365  The answer provided is \"True,\" indicating that...  \n",
       "800         0.412790  The answer \"True\" indicates that the log entry...  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"\\nTop low-trustworthiness log entries:\")\n",
    "low_trust_logs[['log_entry', 'violates_GDPR', 'trustworthiness', 'low_trust_explanation']].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Summary:\n",
      "Total log entries analyzed: 1000\n",
      "GDPR violations (high trustworthiness): 239 (23.90%)\n",
      "Low trustworthiness entries: 8 (0.80%)\n"
     ]
    }
   ],
   "source": [
    "# Calculate and print summary statistics\n",
    "total_logs = len(df)\n",
    "gdpr_violations_count = len(gdpr_violations)\n",
    "low_trust_count = len(low_trust_logs)\n",
    "\n",
    "print(f\"\\nSummary:\")\n",
    "print(f\"Total log entries analyzed: {total_logs}\")\n",
    "print(f\"GDPR violations (high trustworthiness): {gdpr_violations_count} ({gdpr_violations_count/total_logs:.2%})\")\n",
    "print(f\"Low trustworthiness entries: {low_trust_count} ({low_trust_count/total_logs:.2%})\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the results\n",
    "df.to_csv(\"gdpr_analysis_results.csv\", index=False)\n",
    "print(\"\\nResults saved to 'gdpr_analysis_results.csv'\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Most Trustworthy Example:\n",
      "================================================================================\n",
      "\n",
      "Log entry:\n",
      "2024-08-17 17:48:26 INFO: User 'user250@example.com' changed password\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.9998444409832958\n",
      "--------------------------------------------------------------------------------\n",
      "\n",
      "Least Trustworthy Example:\n",
      "================================================================================\n",
      "\n",
      "Log entry:\n",
      "2024-09-09 12:25:00 INFO: Unauthorized access attempt to admin panel from IP:\n",
      "53.125.88.241\n",
      "\n",
      "Violates GDPR: True\n",
      "Trustworthiness score: 0.2336285708561434\n",
      "\n",
      "Low trustworthiness explanation:\n",
      "The answer \"True\" indicates that the log entry violates GDPR by containing or\n",
      "processing personal data without proper safeguards. However, this assessment is\n",
      "not entirely accurate. The log entry contains an IP address, which can be\n",
      "considered personal data under GDPR if it can be linked to an individual.\n",
      "However, the mere presence of an IP address in a log entry does not\n",
      "automatically constitute a GDPR violation. The context of how the IP address is\n",
      "used, stored, and protected is crucial. If the IP address is used solely for\n",
      "security purposes and is adequately protected, it may not violate GDPR.\n",
      "Therefore, the answer \"True\" is not definitively correct without additional\n",
      "context.\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "def display_extreme_trust_examples(df, high_trust_threshold=0.8, low_trust_threshold=0.5):\n",
    "    \"\"\"\n",
    "    Display the highest and lowest trustworthiness examples with wrapped text.\n",
    "    \"\"\"\n",
    "    import textwrap\n",
    "    \n",
    "    # Get single highest trust example\n",
    "    highest_trust = df[df['trustworthiness'] >= high_trust_threshold]\\\n",
    "        .sort_values('trustworthiness', ascending=False)\\\n",
    "        .iloc[0]\n",
    "    \n",
    "    # Get single lowest trust example\n",
    "    lowest_trust = df[(df['violates_GDPR'] == True) & \n",
    "                     (df['trustworthiness'] < low_trust_threshold)]\\\n",
    "        .sort_values('trustworthiness')\\\n",
    "        .iloc[0]\n",
    "    \n",
    "    print(\"\\nMost Trustworthy Example:\")\n",
    "    print(\"=\" * 80)\n",
    "    print(\"\\nLog entry:\")\n",
    "    print(textwrap.fill(highest_trust['log_entry'], width=80))\n",
    "    print(f\"\\nViolates GDPR: {highest_trust['violates_GDPR']}\")\n",
    "    print(f\"Trustworthiness score: {highest_trust['trustworthiness']}\")\n",
    "    print(\"-\" * 80)\n",
    "    \n",
    "    print(\"\\nLeast Trustworthy Example:\")\n",
    "    print(\"=\" * 80)\n",
    "    print(\"\\nLog entry:\")\n",
    "    print(textwrap.fill(lowest_trust['log_entry'], width=80))\n",
    "    print(f\"\\nViolates GDPR: {lowest_trust['violates_GDPR']}\")\n",
    "    print(f\"Trustworthiness score: {lowest_trust['trustworthiness']}\")\n",
    "    if lowest_trust['low_trust_explanation']:\n",
    "        print(\"\\nLow trustworthiness explanation:\")\n",
    "        print(textwrap.fill(lowest_trust['low_trust_explanation'], width=80))\n",
    "    print(\"-\" * 80)\n",
    "\n",
    "# Call the function\n",
    "display_extreme_trust_examples(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mturk-env",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}