Preset Multi-Kerangka: Menganonimkan GDPR, HIPAA...

Preset Multi-Kerangka: Menganonimkan GDPR, HIPAA, dan CCPA Serentak

Organisasi penjagaan kesihatan global mungkin tertakluk kepada:

• GDPR (UE): Melindungi data warga EU
• HIPAA (AS): Melindungi maklumat kesihatan yang dilindungi
• CCPA (California): Melindungi privasi pengguna California

Setiap peraturan mempunyai definisi PII yang sedikit berbeza, keharusan penganoniman, dan senarai "pengecam quasi" yang berbeza.

GDPR Recital 26: Anonymized data is data which does not relate to an identified or identifiable person. HIPAA Safe Harbor: Remove 18 specific identifiers (nama, SSN, tarikh lahir, dll). CCPA § 1798.100: Consumers have the right to deletion of personal information.

Perbezaan Dalam Takrifan PII

Membina Preset Multi-Kerangka

Langkah 1: Tentukan Entiti per Kerangka

FRAMEWORK_ENTITIES = {
  "GDPR": {
    "required": [
      "PERSON",
      "EMAIL_ADDRESS",
      "PHONE_NUMBER",
      "ADDRESS",
      "SSN",
      "CREDIT_CARD",
      "DATE_OF_BIRTH",
      "IP_ADDRESS"
    ],
    "quasi_identifiers": ["COMPANY_NAME", "JOB_TITLE", "LOCATION"]
  },
  "HIPAA": {
    "required": [
      "PERSON",
      "SSN",
      "MEDICAL_RECORD_NUMBER",
      "DATE_OF_BIRTH",
      "PHONE_NUMBER",
      "FAX_NUMBER",
      "EMAIL_ADDRESS",
      "ADDRESS",
      "BIOMETRIC_ID"
    ],
    "quasi_identifiers": ["FACILITY_NAME", "PROVIDER_ID", "HEALTH_PLAN_ID"]
  },
  "CCPA": {
    "required": [
      "PERSON",
      "EMAIL_ADDRESS",
      "ADDRESS",
      "PHONE_NUMBER",
      "COOKIES",
      "IP_ADDRESS",
      "DEVICE_ID",
      "BIOMETRIC_DATA"
    ],
    "quasi_identifiers": ["PURCHASE_HISTORY", "BROWSING_HISTORY", "LOCATION"]
  }
}

Langkah 2: Tentukan Operator Kerangka

FRAMEWORK_OPERATORS = {
  "GDPR": {
    # GDPR memerlukan penganoniman sejati (tidak dapat diubah kembali)
    "PERSON": OperatorConfig("redact"),  # Buang sepenuhnya
    "EMAIL_ADDRESS": OperatorConfig("redact"),
    "SSN": OperatorConfig("redact"),
    # Pengecam kuasi boleh diganti (pseudonimisasi)
    "COMPANY_NAME": OperatorConfig("replace", params={"new_value": "[COMPANY]"}),
    "JOB_TITLE": OperatorConfig("replace", params={"new_value": "[JOB]"})
  },
  "HIPAA": {
    # HIPAA Safe Harbor: buang 18 pengecam
    "PERSON": OperatorConfig("redact"),
    "MEDICAL_RECORD_NUMBER": OperatorConfig("redact"),
    "DATE_OF_BIRTH": OperatorConfig("redact"),
    "PHONE_NUMBER": OperatorConfig("redact"),
    # Pengecam kuasi boleh disimpan atau diganti
    "FACILITY_NAME": OperatorConfig("replace", params={"new_value": "[FACILITY]"})
  },
  "CCPA": {
    # CCPA: pengguna berhak untuk pemadaman
    "PERSON": OperatorConfig("redact"),
    "EMAIL_ADDRESS": OperatorConfig("redact"),
    "COOKIES": OperatorConfig("redact"),
    "DEVICE_ID": OperatorConfig("redact"),
    "BROWSING_HISTORY": OperatorConfig("redact")
  }
}

Langkah 3: Buat Preset Gabungan

def create_multi_framework_preset(document, frameworks=["GDPR", "HIPAA", "CCPA"]):
  analyzer = AnalyzerEngine()
  anonymizer = AnonymizerEngine()
  
  results = analyzer.analyze(
    text=document["content"],
    language="en",
    entities=list(set(
      item for framework in frameworks
      for item in (FRAMEWORK_ENTITIES[framework]["required"] +
                  FRAMEWORK_ENTITIES[framework]["quasi_identifiers"])
    ))
  )
  
  # Untuk setiap kerangka, terapkan operator khususnya
  anonymized_versions = {}
  
  for framework in frameworks:
    framework_operators = FRAMEWORK_OPERATORS[framework]
    
    anonymized = anonymizer.anonymize(
      text=document["content"],
      analyzer_results=results,
      operators={
        entity.entity_type: framework_operators.get(
          entity.entity_type,
          OperatorConfig("replace", params={"new_value": f"[{entity.entity_type}]"})
        )
        for entity in results
      }
    )
    
    anonymized_versions[framework] = anonymized
  
  # Keluarkan versi paling ketat (GDPR redacts semua)
  return anonymized_versions["GDPR"]

Versi Peraturan yang Berbeza untuk Tujuan Berbeza

document = {
  "content": "Pesakit John Smith (DOB 1980-05-15, SSN 123-45-6789, MRN 5731) dirawat di Rumah Sakit Cleveland Clinic."
}

versions = create_multi_framework_preset(document)

print(f"GDPR:  {versions['GDPR']}")
# GDPR: Pesakit [REDACTED] (DOB [REDACTED], SSN [REDACTED], MRN [REDACTED]) dirawat di [FACILITY].

print(f"HIPAA: {versions['HIPAA']}")
# HIPAA: Pesakit [REDACTED] (DOB [REDACTED], SSN [REDACTED], MRN [REDACTED]) dirawat di [FACILITY].

print(f"CCPA:  {versions['CCPA']}")
# CCPA: Pesakit [REDACTED] (DOB [REDACTED], SSN [REDACTED], MRN [REDACTED]) dirawat di [FACILITY].

Kesimpulan: Preset multi-kerangka memastikan dokumen memenuhi kepatuhan semua peraturan tanpa meninggalkan PII di bawah standard apa pun.