Microsoft製のPII検出・匿名化SDK「Presidio」でテキストの個人情報をマスクしてみる

presidio-analyzer 2.2.362
presidio-anonymizer 2.2.362

$ uv pip install presidio_analyzer presidio_anonymizer
$ python -m spacy download en_core_web_lg

from presidio_analyzer import AnalyzerEngine
from presidio_anonymizer import AnonymizerEngine

text = (
    "My name is David Johnson and my email is david.johnson@example.com. "
    "My phone number is (212) 555-1234 and my credit card is 4095-2609-9393-4932."
)

analyzer = AnalyzerEngine()
anonymizer = AnonymizerEngine()

# 1. PII を検出する
results = analyzer.analyze(text=text, language="en")
print("=== 検出結果 ===")
for r in results:
    print(f"{r.entity_type:<16} start={r.start:<3} end={r.end:<3} score={r.score:.2f} value={text[r.start:r.end]!r}")

# 2. 検出した PII を匿名化する
anonymized = anonymizer.anonymize(text=text, analyzer_results=results)
print("\n=== 匿名化後のテキスト ===")
print(anonymized.text)

$ python basic.py
=== 検出結果 ===
EMAIL_ADDRESS    start=41  end=66  score=1.00 value='david.johnson@example.com'
CREDIT_CARD      start=124 end=143 score=1.00 value='4095-2609-9393-4932'
PERSON           start=11  end=24  score=0.85 value='David Johnson'
PHONE_NUMBER     start=87  end=101 score=0.75 value='(212) 555-1234'
URL              start=41  end=49  score=0.50 value='david.jo'
URL              start=55  end=66  score=0.50 value='example.com'

=== 匿名化後のテキスト ===
My name is <PERSON> and my email is <EMAIL_ADDRESS>. My phone number is <PHONE_NUMBER> and my credit card is <CREDIT_CARD>.

from presidio_analyzer import AnalyzerEngine
from presidio_anonymizer import AnonymizerEngine
from presidio_anonymizer.entities import OperatorConfig

text = (
    "My name is David Johnson and my email is david.johnson@example.com. "
    "My phone number is (212) 555-1234 and my credit card is 4095-2609-9393-4932."
)

analyzer = AnalyzerEngine()
anonymizer = AnonymizerEngine()

results = analyzer.analyze(text=text, language="en")

operators = {
    # 既定: 検出値を <REDACTED> に置換
    "DEFAULT": OperatorConfig("replace", {"new_value": "<REDACTED>"}),
    # 人名: 完全に削除
    "PERSON": OperatorConfig("redact", {}),
    # 電話番号: 末尾10文字を * でマスク
    "PHONE_NUMBER": OperatorConfig(
        "mask", {"masking_char": "*", "chars_to_mask": 10, "from_end": True}
    ),
    # クレジットカード: SHA-256 でハッシュ化（salt を固定して再現性を確保）
    "CREDIT_CARD": OperatorConfig(
        "hash", {"hash_type": "sha256", "salt": "presidio-demo-fixed-salt-sample"}
    ),
}

anonymized = anonymizer.anonymize(
    text=text, analyzer_results=results, operators=operators
)
print(anonymized.text)

$ python operators.py
My name is  and my email is <REDACTED>. My phone number is (212********** and my credit card is 498b7ea434e932ed06ac0ca8bd0a23e8ae9f3004c5c0f8ace51324c52998c705.

from presidio_analyzer import AnalyzerEngine, Pattern, PatternRecognizer
from presidio_anonymizer import AnonymizerEngine
from presidio_anonymizer.entities import OperatorConfig

text = "私のマイナンバーは 1234-5678-9012 です。連絡先は test@example.com です。"

# マイナンバー（個人番号: 12桁）を検出するカスタム Recognizer
my_number_pattern = Pattern(
    name="jp_my_number",
    regex=r"\b\d{4}-?\d{4}-?\d{4}\b",
    score=0.5,
)
my_number_recognizer = PatternRecognizer(
    supported_entity="JP_MY_NUMBER",
    patterns=[my_number_pattern],
    context=["マイナンバー", "個人番号", "マイナンバーカード"],
)

analyzer = AnalyzerEngine()
analyzer.registry.add_recognizer(my_number_recognizer)

results = analyzer.analyze(text=text, language="en")
print("=== 検出結果 ===")
for r in results:
    print(f"{r.entity_type:<16} score={r.score:.2f} value={text[r.start:r.end]!r}")

anonymizer = AnonymizerEngine()
anonymized = anonymizer.anonymize(
    text=text,
    analyzer_results=results,
    operators={"JP_MY_NUMBER": OperatorConfig("replace", {"new_value": "<マイナンバー>"})},
)
print("\n=== 匿名化後のテキスト ===")
print(anonymized.text)

$ python custom_recognizer.py
=== 検出結果 ===
EMAIL_ADDRESS    score=1.00 value='test@example.com'
JP_MY_NUMBER     score=0.85 value='1234-5678-9012'
URL              score=0.50 value='example.com'

=== 匿名化後のテキスト ===
私のマイナンバーは <マイナンバー> です。連絡先は <EMAIL_ADDRESS> です。