File size: 34,673 Bytes
73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e 1dd1348 73fba0e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 |
import gradio as gr
import re
import json
import torch
from transformers import pipeline, AutoModelForTokenClassification, AutoTokenizer
import faker
from typing import List, Dict, Any, Optional
import pandas as pd
class EnhancedPiiProtectionPipeline:
"""
A comprehensive PII protection pipeline that:
1. Uses regex for all detectable patterns first
2. Uses multiple custom NER models for remaining detection
3. Provides three protection methods: labeling, masking, and synthesis
4. Handles general, Indian-specific, address, and medical contexts
"""
def __init__(
self,
main_model_name: str = "Kashish-jain/pii-protection-model",
medical_model_name: str = "Kashish-jain/pii-protection-medical",
use_medical_model: bool = False
):
"""
Initialize the comprehensive PII protection pipeline.
Args:
main_model_name: HuggingFace model name or path for the main PII model
medical_model_name: HuggingFace model name for the medical NER model
use_medical_model: Whether to load and use the medical model
"""
# Main model
self.main_tokenizer = AutoTokenizer.from_pretrained(main_model_name)
self.main_model = pipeline("ner", model=main_model_name, tokenizer=self.main_tokenizer, aggregation_strategy="simple")
# Address-specific model - implementation simplified
self.address_model = self.main_model # Fallback to main model for simplicity
# Medical model
self.use_medical_model = use_medical_model
self.medical_model = None
self.medical_tokenizer = None
if use_medical_model and medical_model_name:
try:
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.device = device
self.medical_tokenizer = AutoTokenizer.from_pretrained(medical_model_name)
self.medical_model = pipeline(
"ner",
model=medical_model_name,
tokenizer=self.medical_tokenizer,
aggregation_strategy="simple",
device=0 if torch.cuda.is_available() else -1
)
print(f"Medical model '{medical_model_name}' loaded successfully")
except Exception as e:
print(f"Warning: Could not load medical model. Error: {str(e)}")
self.use_medical_model = False
self.faker = faker.Faker('en_IN')
# Set up regex patterns for common PII entities - IMPROVED PATTERNS
self.regex_patterns = {
# Phone numbers - Fixed to prevent partial matches
'PHONENUMBER': r'(?<!\w)(?:\+91[\-\s]?[789]\d{9}|(?:\+91[\-\s]?)?\d{3}[\-\.\s]?\d{3}[\-\.\s]?\d{4}|(?:\d{3}[\-\s]?){2}\d{4})(?!\d)',
# Email
'EMAIL': r'(?<!\w)[a-zA-Z0-9._%+\-]+@[a-zA-Z0-9.\-]+\.[a-zA-Z]{2,}(?!\w)',
# IP addresses
'IPV4': r'(?<!\w)(?:(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.){3}(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)(?!\w)',
# Credit cards
'CREDITCARDNUMBER': r'(?<!\w)(?:4\d{12}(?:\d{3})?|5[1-5]\d{14}|6(?:011|5\d{2})\d{12}|3[47]\d{13}|3(?:0[0-5]|[68]\d)\d{11}|(?:2131|1800|35\d{3})\d{11})(?!\w)',
# PAN (Indian Permanent Account Number)
'PAN': r'(?<!\w)[A-Z]{5}[0-9]{4}[A-Z](?!\w)',
# Aadhar (Indian ID)
'AADHAR': r'(?<!\w)(?:\d{4}\s\d{4}\s\d{4}|\d{12})(?!\d)',
# Passport
'PASSPORT': r'(?<!\w)[A-Z]{1,2}\d{7}(?!\w)',
# URL
'URL': r'(?<!\w)https?://(?:www\.)?[-a-zA-Z0-9@:%._\+~#=]{1,256}\.[a-zA-Z0-9()]{1,6}\b(?:[-a-zA-Z0-9()@:%_\+.~#?&//=]*)(?!\w)',
# Dates
'DOB': r'(?<!\w)(?:0[1-9]|[12][0-9]|3[01])[/\-\.](?:0[1-9]|1[0-2])[/\-\.](?:19|20)\d{2}(?!\w)',
# PINCODE
'PINCODE': r'(?<!\w)(?:PIN[\s-]*)?\d{6}(?!\d)',
# Bank account & IBAN
'ACCOUNTNUMBER': r'(?<!\w)(?:A/C|Account|ACC)(?:ount)?\s*(?:Number|No|#)?[:\s-]*(\d{9,17})(?!\d)',
'IBAN_CODE': r'(?<!\w)(?:IBAN|International Bank Account Number)?[:\s]*[A-Z]{2}\d{2}[A-Z0-9]{4}[0-9]{7}(?:[0-9]{0,16})(?!\w)',
# Social Security Number (US)
'SSN': r'(?<!\w)\d{3}[-\s]?\d{2}[-\s]?\d{4}(?!\w)',
# Driver's License (simplified)
'DRIVER_LICENSE': r'(?<!\w)(?:[A-Z]{1,2}-\d{5,8}|\d{7,9}|[A-Z]\d{3}-\d{4}-\d{4}|\d{3}-\d{2}-\d{4})(?!\w)'
}
# Medical entity regex patterns - ENHANCED to only capture the value part, not label
self.medical_regex_patterns = {
'DOCTORNAME': r'(?:Dr\.?|Doctor)\s+([A-Z][a-z]+(?:\s+[A-Z][a-z]+)?)',
'PATIENTID': r'(?:Patient\s+ID|ID|MRN)[\s-]*[:]\s*([A-Z0-9]{5,12})', # Modified to use a capture group
'MEDICALID': r'(?:Medical\s+Record|MRN|Patient\s+ID)[\s-]*[:]\s*([A-Z0-9]{4,15})', # Modified to use a capture group
}
# Separated measurements with capture groups to get just the values, not labels
self.measurement_patterns = {
# Height with capture group for just the measurement value
'HEIGHT': r'(?:Height|Ht)[\s-]*[:]\s*((?:\d{1,2}\'\s*(?:\d{1,2}\")?|\d{3}\s*cm|\d{1,2}\.\d{1,2}\s*m))',
# Weight with capture group for just the measurement value
'WEIGHT': r'(?:Weight|Wt)[\s-]*[:]\s*((?:\d{1,3}(?:\.\d{1,2})?\s*(?:kg|lbs?|pounds?|kilograms?)))',
# Blood group/type with separate regex for the value only
'BLOOD_TYPE': r'(?:Blood\s+[Tt]ype|Blood\s+[Gg]roup)[\s-]*[:]\s*((?:A|B|AB|O)[+-])',
}
# Standalone measurement patterns (no labels)
self.standalone_medical_patterns = {
'HEIGHT_STANDALONE': r'(?<!\w)(?:\d{1,2}\'\s*\d{1,2}\"|\d{1,2}\'\d{1,2}\"|\d{1,2}\'|\d{3}\s*cm|\d{1,2}\.\d{1,2}\s*m)(?!\w)',
'WEIGHT_STANDALONE': r'(?<!\w)(?:\d{1,3}(?:\.\d{1,2})?\s*(?:kg|lbs?|pounds?|kilograms?))(?!\w)',
'BLOOD_TYPE_STANDALONE': r'(?<!\w)(?:A|B|AB|O)[+-](?!\w)'
}
# Combine all regex patterns
self.all_regex_patterns = {
**self.regex_patterns,
**self.medical_regex_patterns,
**self.measurement_patterns,
**self.standalone_medical_patterns
}
def regex_detection(self, text: str) -> List[Dict[str, Any]]:
"""Detect PII using regex patterns with improved capture groups."""
entities = []
for entity_type, pattern in self.all_regex_patterns.items():
for match in re.finditer(pattern, text, re.IGNORECASE):
# For patterns with capture groups, use the first group if it exists
if match.groups() and match.group(1):
# For labeled patterns with capture groups (e.g., "Height: 5'6"")
captured_text = match.group(1)
# Calculate start/end positions for the captured group
start = match.start(1)
end = match.end(1)
else:
# For patterns without capture groups or standalone measurements
captured_text = match.group(0)
start = match.start(0)
end = match.end(0)
# Handle standalone height/weight by renaming them
if entity_type == 'HEIGHT_STANDALONE':
entity_type = 'HEIGHT'
elif entity_type == 'WEIGHT_STANDALONE':
entity_type = 'WEIGHT'
elif entity_type == 'BLOOD_TYPE_STANDALONE':
entity_type = 'BLOOD_TYPE'
entities.append({
"text": captured_text,
"label": entity_type,
"start": start,
"end": end,
"score": 0.95, # High confidence for regex matches
"_original_text": text # Store original text for context
})
return entities
def ner_detection(self, text: str, model_type: str = "main") -> List[Dict[str, Any]]:
"""
Detect PII using NER models
Args:
text: Text to analyze
model_type: Type of model to use ("main", "medical")
"""
if model_type == "medical" and not self.use_medical_model:
return []
model = self.medical_model if model_type == "medical" else self.main_model
try:
results = model(text)
# Convert to standard format
entities = []
for result in results:
# Skip low confidence predictions
if result.get('score', 0) < 0.5:
continue
# Clean entity type
entity_type = result.get('entity_group', result.get('entity', '')).replace('B-', '').replace('I-', '')
entities.append({
"text": result.get('word', text[result['start']:result['end']]),
"label": entity_type,
"start": result['start'],
"end": result['end'],
"score": result.get('score', 0.7),
"_original_text": text # Store original text for context
})
return entities
except Exception as e:
print(f"Error with NER detection: {str(e)}")
return []
def merge_entities(self, entities: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
"""Merge adjacent entities of the same or related types that likely form a single entity"""
if not entities:
return []
# Sort entities by start position
entities.sort(key=lambda x: x['start'])
merged = []
# Define related entity groups (entities that could be part of the same larger entity)
related_types = {
'NAME': ['FIRSTNAME', 'MIDDLENAME', 'LASTNAME', 'PREFIX'],
'ADDRESS': ['STREET', 'CITY', 'STATE', 'ZIPCODE', 'BUILDINGNUMBER'],
'PHONENUMBER': ['PHONENUMBER'] # Explicitly add PHONENUMBER to prevent merging with other types
}
# Flatten the related types for quick lookup
related_types_flat = {}
for main_type, sub_types in related_types.items():
for sub_type in sub_types:
related_types_flat[sub_type] = main_type
# Helper function to check if two entity types are related
def are_related(type1, type2):
# Same type is related
if type1 == type2:
return True
# Prevent merging PHONENUMBER with other types
if type1 == 'PHONENUMBER' or type2 == 'PHONENUMBER':
return type1 == type2
# Check if they're in the same group
for group, types in related_types.items():
if type1 in types and type2 in types:
return True
if type1 == group and type2 in types:
return True
if type2 == group and type1 in types:
return True
# Check through the flattened related types
if type1 in related_types_flat and related_types_flat[type1] == type2:
return True
if type2 in related_types_flat and related_types_flat[type2] == type1:
return True
return False
for entity in entities:
if not merged:
merged.append(entity.copy())
continue
last = merged[-1]
# Maximum space between tokens that could be part of the same entity
# For adjacent words, this would typically be 1 (the space)
max_gap = 5
# Check if entities could be part of the same larger entity:
# 1. Same or related entity type
# 2. Within a reasonable distance
# 3. No other complete word between them
if (are_related(entity['label'], last['label']) and
entity['start'] - last['end'] <= max_gap):
# Get the text between the two entities
between_text = entity.get('_original_text', '')[last['end']:entity['start']] \
if '_original_text' in entity and '_original_text' in last \
else ' '
# Only merge if the gap contains just spaces or very simple punctuation
if between_text.strip() in ['', ' ', '.', ',', '-', '_']:
# Create merged entity with all text between start and end
if '_original_text' in entity and '_original_text' in last:
full_text = last['_original_text'][last['start']:entity['end']]
else:
full_text = last['text'] + between_text + entity['text']
last['text'] = full_text
last['end'] = entity['end']
# When merging different entity types, prefer the broader category
if last['label'] in related_types_flat and entity['label'] == related_types_flat[last['label']]:
last['label'] = entity['label']
elif entity['label'] in related_types_flat and last['label'] == related_types_flat[entity['label']]:
# Keep last['label'] as is
pass
last['score'] = max(last.get('score', 0), entity.get('score', 0))
else:
merged.append(entity.copy())
else:
merged.append(entity.copy())
return merged
def remove_overlapping_entities(self, entities: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
"""Remove overlapping entities by keeping the highest scoring one"""
if not entities:
return []
# Sort by start position
entities.sort(key=lambda x: x['start'])
# Identify overlapping entities
non_overlapping = []
i = 0
while i < len(entities):
current = entities[i]
# Find all entities that overlap with the current one
overlapping = [current]
j = i + 1
while j < len(entities) and entities[j]['start'] < current['end']:
overlapping.append(entities[j])
j += 1
# Keep the highest scoring entity from overlapping group
if len(overlapping) > 1:
best_entity = max(overlapping, key=lambda x: x.get('score', 0))
non_overlapping.append(best_entity)
else:
non_overlapping.append(current)
# Move index to start after all overlapping entities
i = j
return non_overlapping
def generate_synthetic_value(self, entity_type: str, original_value: str = None) -> str:
"""Generate realistic synthetic data for PII."""
try:
if entity_type in ['PERSON', 'NAME', 'FIRSTNAME', 'LASTNAME']:
return self.faker.name()
elif entity_type == 'EMAIL':
return self.faker.email()
elif entity_type == 'PHONENUMBER':
return self.faker.phone_number()
elif entity_type == 'PAN':
return self.faker.bothify('?????####?').upper()
elif entity_type == 'AADHAR':
return ' '.join([self.faker.numerify('####') for _ in range(3)])
elif entity_type == 'CREDITCARDNUMBER' or entity_type == 'CREDIT_CARD':
return self.faker.credit_card_number()
elif entity_type == 'ACCOUNTNUMBER' or entity_type == 'IBAN_CODE' or entity_type == 'BANK_NUMBER':
return self.faker.bban()
elif entity_type == 'PASSPORT' or entity_type == 'US_PASSPORT':
return f"{self.faker.random_letter().upper()}{self.faker.random_letter().upper()}{self.faker.numerify('######')}"
elif entity_type == 'DOB' or entity_type == 'DATE_TIME':
return self.faker.date_of_birth(minimum_age=18, maximum_age=90).strftime('%d/%m/%Y')
elif entity_type == 'IPV4' or entity_type == 'IP_ADDRESS':
return self.faker.ipv4()
elif entity_type == 'URL':
return self.faker.url()
elif entity_type == 'PINCODE':
return self.faker.postcode()
elif entity_type == 'CITY' or entity_type == 'LOCATION':
return self.faker.city()
elif entity_type == 'STATE':
return self.faker.state()
elif entity_type == 'SSN' or entity_type == 'US_SSN':
return self.faker.ssn()
elif entity_type == 'DRIVER_LICENSE' or entity_type == 'US_DRIVER_LICENSE':
return self.faker.bothify('?#######')
elif entity_type == 'CRYPTO':
return self.faker.cryptocurrency_code() + self.faker.bothify('??##??##??##??')
# Medical entity generation
elif entity_type == 'DOCTORNAME':
return f"Dr. {self.faker.last_name()}"
elif entity_type == 'PATIENTID' or entity_type == 'MEDICALID':
return self.faker.bothify('PT#######')
elif entity_type == 'HEIGHT':
# Generate a realistic height in feet and inches
feet = self.faker.random_int(min=4, max=6)
inches = self.faker.random_int(min=0, max=11)
return f"{feet}'{inches}\""
elif entity_type == 'WEIGHT':
# Generate a realistic weight in kg
weight = self.faker.random_int(min=45, max=100)
return f"{weight}kg"
elif entity_type == 'BLOOD_TYPE':
# Generate a random blood type
blood_groups = ['A+', 'A-', 'B+', 'B-', 'AB+', 'AB-', 'O+', 'O-']
return self.faker.random_element(blood_groups)
else:
# Fallback for unknown types
return f"[SYNTHETIC_{entity_type}]"
except Exception as e:
print(f"Error generating synthetic value: {str(e)}")
return f"[SYNTHETIC_{entity_type}]"
def process_text(self, text: str, model_type: str = "main", protection_method: str = "replace") -> Dict[str, Any]:
"""
Process text to detect and protect PII
Args:
text: Input text to process
model_type: Type of model to use ("main", "medical")
protection_method: Protection method ("replace", "mask", "synthesize")
Returns:
Dict containing protected text and detected entities
"""
# Step 1: Get entities from regex
regex_entities = self.regex_detection(text)
# Step 2: Get entities from NER model
ner_entities = self.ner_detection(text, model_type)
# Step 3: Combine and process entities
all_entities = regex_entities + ner_entities
merged_entities = self.merge_entities(all_entities)
final_entities = self.remove_overlapping_entities(merged_entities)
# Step 4: Create protected text based on method
protected_text = text
# Sort entities by start position in reverse to avoid index issues when replacing
final_entities_sorted = sorted(final_entities, key=lambda x: x['start'], reverse=True)
if protection_method == "mask":
# Mask with asterisks
for entity in final_entities_sorted:
mask = '*' * len(entity['text'])
protected_text = protected_text[:entity['start']] + mask + protected_text[entity['end']:]
elif protection_method == "synthesize":
# Replace with synthetic values
for entity in final_entities_sorted:
synthetic = self.generate_synthetic_value(entity['label'], entity['text'])
protected_text = protected_text[:entity['start']] + synthetic + protected_text[entity['end']:]
else: # replace (default)
# Replace with entity tags
for entity in final_entities_sorted:
tag = f"[{entity['label']}]"
protected_text = protected_text[:entity['start']] + tag + protected_text[entity['end']:]
# Create findings table
findings = []
for i, entity in enumerate(final_entities):
findings.append({
"index": i,
"entity_type": entity['label'],
"text": entity['text'],
"start": entity['start'],
"end": entity['end'],
"confidence": round(entity.get('score', 1.0), 2)
})
return {
"protected_text": protected_text,
"entities": final_entities,
"findings": findings
}
# Example input text
example_text = """
Hi, my name is John Doe and I'm originally from Delhi.
On 11/10/2024 I visited https://www.google.com and sent an email to abc@gmail.com, from IP 192.168.0.1.
My phone number: +91-1234321216.
"""
medical_example_text = """
Patient name: John Doe
Date of Birth: 05/12/1982
Patient ID: PT789456
Contact: +91-9876543210
Dr. Robert Johnson has prescribed medication penicillin on 12/12/2024.
Blood type: O+, Height: 5'6", Weight: 145kg
"""
# Create Gradio Interface
def process_input(text, model_type, protection_method):
# Initialize pipeline with Hugging Face model paths
main_model_name = "Kashish-jain/pii-protection-model"
medical_model_name = "Kashish-jain/pii-protection-medical"
use_medical = model_type == "medical"
pipeline = EnhancedPiiProtectionPipeline(
main_model_name=main_model_name,
medical_model_name=medical_model_name,
use_medical_model=use_medical
)
# Process the text
result = pipeline.process_text(text, model_type, protection_method)
# Create findings table
if result["findings"]:
df = pd.DataFrame(result["findings"])
df = df.rename(columns={
"index": "#",
"entity_type": "Entity type",
"text": "Text",
"start": "Start",
"end": "End",
"confidence": "Confidence"
})
else:
df = pd.DataFrame(columns=["#", "Entity type", "Text", "Start", "End", "Confidence"])
# Count detected entities by type
if result["findings"]:
entity_counts = df["Entity type"].value_counts().to_dict()
entity_summary = ", ".join([f"{count} {entity}" for entity, count in entity_counts.items()])
else:
entity_summary = "No entities detected"
return result["protected_text"], df, entity_summary
# Update input text based on model type
def update_input_text(model_type):
if model_type == "medical":
return medical_example_text
else:
return example_text
# Custom CSS for a minimalistic, clean design
custom_css = """
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&family=Playfair+Display:wght@400;700&display=swap');
:root {
--font-sans: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, sans-serif;
--font-serif: 'Playfair Display', Georgia, Cambria, 'Times New Roman', Times, serif;
--color-primary: #2563eb;
--color-primary-light: #3b82f6;
--color-primary-dark: #1d4ed8;
--color-secondary: #64748b;
--color-secondary-light: #94a3b8;
--color-background: #00000f;
--color-surface: #f8fafc;
--color-border: #e2e8f0;
--color-text: #1e293b;
--color-text-light: #64748b;
--color-success: #10b981;
--color-warning: #f59e0b;
--color-error: #ef4444;
--shadow-sm: 0 1px 2px 0 rgba(0, 0, 0, 0.05);
--shadow: 0 1px 3px 0 rgba(0, 0, 0, 0.1), 0 1px 2px 0 rgba(0, 0, 0, 0.06);
--shadow-md: 0 4px 6px -1px rgba(0, 0, 0, 0.1), 0 2px 4px -1px rgba(0, 0, 0, 0.06);
--shadow-lg: 0 10px 15px -3px rgba(0, 0, 0, 0.1), 0 4px 6px -2px rgba(0, 0, 0, 0.05);
--radius-sm: 0.25rem;
--radius: 0.375rem;
--radius-md: 0.5rem;
--radius-lg: 0.75rem;
--spacing-1: 0.25rem;
--spacing-2: 0.5rem;
--spacing-3: 0.75rem;
--spacing-4: 1rem;
--spacing-6: 1.5rem;
--spacing-8: 2rem;
--spacing-12: 3rem;
}
body, .gradio-container {
font-family: var(--font-sans);
color: var(--color-text);
background-color: var(--color-background);
line-height: 1.5;
}
/* Typography */
h1, h2, h3 {
font-family: var(--font-serif);
font-weight: 700;
line-height: 1.2;
margin-bottom: var(--spacing-4);
}
h1 {
font-size: 2.25rem;
color: var(--color-text-light);
}
h2 {
font-size: 1.5rem;
color: var(--color-text);
}
h3 {
font-size: 1.25rem;
color: var(--color-text);
}
p {
margin-bottom: var(--spacing-4);
}
/* Layout Components */
.container {
max-width: 1500px;
margin: 0 auto;
padding: var(--spacing-6);
}
.card {
background-color: var(--color-surface);
border-radius: var(--radius);
box-shadow: var(--shadow);
padding: var(--spacing-6);
margin-bottom: var(--spacing-6);
border: 1px solid var(--color-border);
}
/* Form Elements */
.gradio-button.primary {
background-color: var(--color-secondary-light);
color: white;
font-weight: 500;
border-radius: var(--radius);
padding: var(--spacing-3) var(--spacing-6);
transition: all 0.2s ease;
border: none;
box-shadow: var(--shadow);
}
.gradio-button.primary:hover {
background-color: var(--color-secondary);
box-shadow: var(--shadow-md);
transform: translateY(-1px);
}
.gradio-button.primary:active {
transform: translateY(0);
}
.gradio-dropdown, .gradio-textbox, .gradio-textarea {
border-radius: var(--radius);
border: 1px solid var(--color-border);
padding: var(--spacing-3);
background-color: var(--color-background);
transition: border-color 0.2s ease;
}
.gradio-dropdown:focus, .gradio-textbox:focus, .gradio-textarea:focus {
border-color: var(--color-primary-light);
outline: none;
box-shadow: 0 0 0 3px rgba(37, 99, 235, 0.1);
}
/* Tabs */
.gradio-tabs {
margin-bottom: var(--spacing-6);
}
.gradio-tab-button {
padding: var(--spacing-3) var(--spacing-6);
font-weight: 500;
color: var(--color-text-light);
border-bottom: 2px solid transparent;
transition: all 0.2s ease;
}
.gradio-tab-button.selected {
color: var(--color-primary);
border-bottom-color: var(--color-primary);
}
/* Accordion */
.gradio-accordion {
border: 1px solid var(--color-border);
border-radius: var(--radius);
margin-bottom: var(--spacing-6);
overflow: hidden;
}
.gradio-accordion-header {
padding: var(--spacing-4);
font-weight: 500;
background-color: var(--color-surface);
border-bottom: 1px solid var(--color-border);
cursor: pointer;
}
.gradio-accordion-content {
padding: var(--spacing-4);
background-color: var(--color-background);
}
/* Table */
table {
width: 100%;
border-collapse: collapse;
margin-bottom: var(--spacing-6);
}
th {
background-color: var(--color-surface);
padding: var(--spacing-3) var(--spacing-4);
text-align: left;
font-weight: 600;
color: var(--color-text);
border-bottom: 2px solid var(--color-border);
}
td {
padding: var(--spacing-3) var(--spacing-4);
border-bottom: 1px solid var(--color-border);
}
/* Dark mode support */
@media (prefers-color-scheme: dark) {
:root {
--color-background: #0f172a;
--color-surface: #1e293b;
--color-border: #334155;
--color-text: #f8fafc;
--color-text-light: #cbd5e1;
}
}
/* Custom components */
.entity-badge {
display: inline-block;
padding: 0.25rem 0.5rem;
border-radius: 9999px;
font-size: 0.75rem;
font-weight: 500;
background-color: var(--color-primary-light);
color: white;
margin-right: 0.5rem;
margin-bottom: 0.5rem;
}
.summary-container {
background-color: var(--color-surface);
border-radius: var(--radius);
padding: var(--spacing-4);
margin-bottom: var(--spacing-6);
border: 1px solid var(--color-border);
}
.icon-text {
display: flex;
align-items: center;
gap: var(--spacing-2);
}
.icon-text svg {
width: 1.25rem;
height: 1.25rem;
color: var(--color-primary);
}
/* Responsive adjustments */
@media (max-width: 768px) {
.container {
padding: var(--spacing-4);
}
h1 {
font-size: 1.75rem;
}
.card {
padding: var(--spacing-4);
}
}
"""
# Create the Gradio interface with enhanced styling
with gr.Blocks(css=custom_css, theme=gr.themes.Base()) as demo:
# Header section
with gr.Column(elem_classes="container"):
gr.Markdown("""
# 🛡️ PII Protection Tool
Detect, protect and de-identify personally identifiable information.
""")
# Main content area
with gr.Column(elem_classes="card"):
# Configuration section
with gr.Row():
with gr.Column(scale=1):
model_dropdown = gr.Dropdown(
choices=[
("General Purpose", "main"),
("Medical Context", "medical")
],
value="main",
label="Model Type",
elem_classes="form-control"
)
with gr.Column(scale=1):
protection_dropdown = gr.Dropdown(
choices=[
("Replace with Tags", "replace"),
("Mask with Asterisks", "mask"),
("Generate Synthetic Data", "synthesize")
],
value="replace",
label="Protection Method",
elem_classes="form-control"
)
# Divider
gr.Markdown("---")
# Input/Output section
with gr.Row():
# Input column
with gr.Column():
gr.Markdown("### Input Text")
input_text = gr.TextArea(
label="",
value=example_text,
lines=10,
elem_classes="text-input"
)
# Output column
with gr.Column():
gr.Markdown("### Protected Output")
output_text = gr.TextArea(
label="",
lines=10,
elem_classes="text-output"
)
# Summary section
with gr.Column(elem_classes="summary-container"):
gr.Markdown("### Entity Summary")
entity_summary = gr.Textbox(
label="",
interactive=False,
elem_classes="entity-summary"
)
# Action button
submit_btn = gr.Button(
"Process Text",
variant="primary",
elem_classes="submit-button"
)
# Findings section
with gr.Column(elem_classes="card"):
gr.Markdown("### Detected Entities")
findings_table = gr.DataFrame(
headers=["#", "Entity type", "Text", "Start", "End", "Confidence"],
elem_classes="findings-table"
)
# Help section
with gr.Accordion("Help & Information", open=False, elem_classes="help-accordion"):
gr.Markdown("""
#### De-identification Methods
- **Replace with Tags**: Replaces each detected entity with its entity type tag (e.g., [NAME])
- **Mask with Asterisks**: Replaces each detected entity with asterisks (*)
- **Generate Synthetic Data**: Replaces each detected entity with realistic synthetic data
#### Model Types
- **General Purpose**: Optimized for common PII elements
- **Medical Context**: Enhanced detection for healthcare-related PII
#### Entity Types Detected
- **Personal**: NAME, EMAIL, PHONENUMBER, DOB
- **Financial**: CREDITCARDNUMBER, ACCOUNTNUMBER, PAN, IBAN_CODE, SSN
- **Location**: ADDRESS, CITY, STATE, PINCODE, IPV4
- **Medical**: DOCTORNAME, PATIENTID, MEDICALID
- **Other**: URL, PASSPORT, DRIVER_LICENSE
""")
# Set up event handlers
submit_btn.click(
fn=process_input,
inputs=[input_text, model_dropdown, protection_dropdown],
outputs=[output_text, findings_table, entity_summary]
)
model_dropdown.change(
fn=update_input_text,
inputs=[model_dropdown],
outputs=[input_text]
)
# Launch the app
if __name__ == "__main__":
demo.launch()
|