@doccloak/core 0.8.0

package/rules/it.json

@@ -0,0 +1,116 @@
+{
+  "region": "it",
+  "rules": [
+    {
+      "id": "regex:it:codice_fiscale",
+      "entityType": "SSN",
+      "pattern": "\\b[A-Z]{6}\\d{2}[A-EHLMPRST]\\d{2}[A-Z]\\d{3}[A-Z]\\b",
+      "flags": "gi",
+      "confidence": 0.9,
+      "domains": [
+        "identity",
+        "hr",
+        "medical"
+      ],
+      "description": "Italian fiscal code (Codice Fiscale, 16 alphanumeric characters)",
+      "examples": [
+        "RSSMRA85M01H501Q"
+      ],
+      "validate": "codiceFiscale"
+    },
+    {
+      "id": "regex:it:partita_iva",
+      "entityType": "SSN",
+      "pattern": "\\bIT\\d{11}\\b",
+      "flags": "gi",
+      "confidence": 0.75,
+      "domains": [
+        "financial",
+        "legal"
+      ],
+      "description": "Italian VAT number (Partita IVA, IT + 11 digits)",
+      "examples": [
+        "IT12345678903"
+      ],
+      "validate": "partitaIva"
+    },
+    {
+      "id": "regex:it:phone_mobile",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?39[\\s-]?)?3[0-9]{2}[\\s-]?\\d{3}[\\s-]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Italian mobile phone number (3xx prefix)",
+      "examples": [
+        "+39 345 123 4567",
+        "3451234567"
+      ]
+    },
+    {
+      "id": "regex:it:phone_landline",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?39[\\s-]?)?0\\d{1,3}[\\s-]?\\d{4,8}\\b",
+      "flags": "g",
+      "confidence": 0.7,
+      "domains": [
+        "contact"
+      ],
+      "description": "Italian landline phone number (0x prefix)",
+      "examples": [
+        "+39 06 12345678",
+        "02 12345678"
+      ]
+    },
+    {
+      "id": "regex:it:street",
+      "entityType": "ADDRESS",
+      "pattern": "(?:via|viale|v\\.le|piazza|p\\.zza|piazzale|corso|c\\.so|largo|vicolo|borgo|salita)\\s+[\\p{L}][\\p{L}\\s'.,-]+\\s*(?:,?\\s*n[°.]?\\s*)?\\d{1,5}[\\p{L}]?",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "contact"
+      ],
+      "description": "Italian street address (Via/Piazza/Corso + name + number)",
+      "examples": [
+        "Via Roma, 42",
+        "Piazza Navona 1",
+        "Corso Vittorio Emanuele II, n. 23"
+      ]
+    },
+    {
+      "id": "regex:it:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{5}\\s+[\\p{L}][\\p{L}\\s-]+\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Italian postal code (CAP, 5 digits) followed by city name",
+      "examples": [
+        "00100 Roma",
+        "20121 Milano"
+      ],
+      "falsePositiveNotes": "5-digit postal codes overlap with US ZIP codes and other number formats"
+    },
+    {
+      "id": "regex:it:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s&.']+(?:\\s+)?(?:S\\.?r\\.?l\\.?|S\\.?p\\.?A\\.?|S\\.?a\\.?s\\.?|S\\.?n\\.?c\\.?|S\\.?a\\.?p\\.?a\\.?)\\b",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Italian company name with legal form (S.r.l., S.p.A., S.a.s., S.n.c., S.a.p.a.)",
+      "examples": [
+        "Fiat Chrysler S.p.A.",
+        "Ristorante Da Mario S.r.l."
+      ]
+    }
+  ]
+}

package/rules/jp.json

@@ -0,0 +1,136 @@
+{
+  "region": "jp",
+  "rules": [
+    {
+      "id": "regex:jp:my_number",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{4}\\s?\\d{4}\\s?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.7,
+      "domains": [
+        "identity"
+      ],
+      "description": "Japanese My Number (individual number, 12 digits)",
+      "examples": [
+        "1234 5678 9012"
+      ]
+    },
+    {
+      "id": "regex:jp:passport",
+      "entityType": "SSN",
+      "pattern": "\\b[A-Z]{2}\\d{7}\\b",
+      "flags": "g",
+      "confidence": 0.7,
+      "domains": [
+        "identity"
+      ],
+      "description": "Japanese passport number (2 letters + 7 digits)",
+      "examples": [
+        "TK1234567"
+      ]
+    },
+    {
+      "id": "regex:jp:corporate_number",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{13}\\b",
+      "flags": "g",
+      "confidence": 0.5,
+      "domains": [
+        "identity",
+        "financial"
+      ],
+      "description": "Japanese corporate number (houjin bangou, 13 digits)",
+      "examples": [
+        "1234567890123"
+      ]
+    },
+    {
+      "id": "regex:jp:phone",
+      "entityType": "PHONE",
+      "pattern": "\\b0\\d{1,4}[\\s-]?\\d{1,4}[\\s-]?\\d{3,4}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Japanese phone number (0X-XXXX-XXXX, variable area code length)",
+      "examples": [
+        "03-1234-5678",
+        "090-1234-5678",
+        "0120-123-456"
+      ]
+    },
+    {
+      "id": "regex:jp:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{3}-?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.65,
+      "domains": [
+        "contact"
+      ],
+      "description": "Japanese postal code (XXX-XXXX)",
+      "examples": [
+        "100-0001",
+        "1600023"
+      ]
+    },
+    {
+      "id": "regex:jp:currency_kanji",
+      "entityType": "CURRENCY",
+      "pattern": "[一二三四五六七八九十百千万億兆]+\\s*円",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Japanese amount in kanji numerals + yen sign (e.g., 八千五百円)",
+      "examples": [
+        "八千五百円"
+      ]
+    },
+    {
+      "id": "regex:jp:address",
+      "entityType": "ADDRESS",
+      "pattern": "[\\u4E00-\\u9FFF]+[都道府県][\\u4E00-\\u9FFF]+[市区町村郡][\\u4E00-\\u9FFF\\d-]+",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "Japanese address (prefecture + city/ward + district)",
+      "examples": [
+        "東京都千代田区丸の内1-1-1"
+      ]
+    },
+    {
+      "id": "regex:jp:company_prefix",
+      "entityType": "COMPANY",
+      "pattern": "(?:株式会社|有限会社|合同会社|合名会社|合資会社)\\s*[\\u4E00-\\u9FFF\\p{L}\\s]+",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "general"
+      ],
+      "description": "Japanese company with legal prefix (kabushiki gaisha etc.)",
+      "examples": [
+        "株式会社トヨタ自動車",
+        "有限会社田中商店"
+      ]
+    },
+    {
+      "id": "regex:jp:company_suffix",
+      "entityType": "COMPANY",
+      "pattern": "[\\u4E00-\\u9FFF\\p{L}][\\u4E00-\\u9FFF\\p{L}\\s]+(?:株式会社|有限会社|合同会社)",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "general"
+      ],
+      "description": "Japanese company with legal suffix (name + kabushiki gaisha)",
+      "examples": [
+        "トヨタ自動車株式会社"
+      ]
+    }
+  ]
+}

package/rules/nl.json

@@ -0,0 +1,99 @@
+{
+  "region": "nl",
+  "rules": [
+    {
+      "id": "regex:nl:bsn",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{9}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "identity",
+        "hr"
+      ],
+      "description": "Dutch citizen service number (BSN, 9 digits with elfproef validation)",
+      "examples": [
+        "123456782"
+      ],
+      "falsePositiveNotes": "9 bare digits match many formats; elfproef validation essential",
+      "validate": "bsn"
+    },
+    {
+      "id": "regex:nl:phone_mobile",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?31[\\s-]?|0)6[\\s-]?\\d{4}[\\s-]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "Dutch mobile phone number (06 prefix)",
+      "examples": [
+        "+31 6 1234 5678",
+        "06 12345678"
+      ]
+    },
+    {
+      "id": "regex:nl:phone_landline",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?31[\\s-]?|0)[1-5]\\d[\\s-]?\\d{3}[\\s-]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.75,
+      "domains": [
+        "contact"
+      ],
+      "description": "Dutch landline phone number (area code prefix)",
+      "examples": [
+        "+31 20 123 4567",
+        "020 1234567"
+      ]
+    },
+    {
+      "id": "regex:nl:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{4}\\s?[A-Z]{2}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Dutch postal code (4 digits + 2 letters)",
+      "examples": [
+        "1012 AB",
+        "3511XA"
+      ]
+    },
+    {
+      "id": "regex:nl:street",
+      "entityType": "ADDRESS",
+      "pattern": "[\\p{L}][\\p{L}\\s'-]*(?:straat|weg|laan|plein|gracht|kade|singel|dijk|steeg|dreef|hof|markt)\\s+\\d{1,5}[\\p{L}]?",
+      "flags": "giu",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "Dutch street address (name + straat/weg/laan/gracht + number)",
+      "examples": [
+        "Keizersgracht 123",
+        "Dorpsstraat 45a",
+        "Marktplein 7"
+      ]
+    },
+    {
+      "id": "regex:nl:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s&.']+(?:\\s+)?(?:B\\.?V\\.?|N\\.?V\\.?|V\\.?O\\.?F\\.?|C\\.?V\\.?)\\b",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Dutch company name with legal form (B.V., N.V., V.O.F., C.V.)",
+      "examples": [
+        "Philips N.V.",
+        "Shell Nederland B.V."
+      ]
+    }
+  ]
+}

package/rules/no.json

@@ -0,0 +1,132 @@
+{
+  "region": "no",
+  "rules": [
+    {
+      "id": "regex:no:fodselsnummer",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{6}\\s?\\d{5}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "identity",
+        "hr",
+        "medical"
+      ],
+      "description": "Norwegian fødselsnummer (birth number, 11 digits, DDMMYY + 5)",
+      "examples": [
+        "01019912368"
+      ],
+      "falsePositiveNotes": "11-digit sequences are common in other contexts",
+      "validate": "fodselsnummer"
+    },
+    {
+      "id": "regex:no:d_nummer",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{6}\\s?\\d{5}\\b",
+      "flags": "g",
+      "confidence": 0.4,
+      "domains": [
+        "identity"
+      ],
+      "description": "Norwegian D-nummer (temporary ID for foreigners, 11 digits, day +40)",
+      "examples": [
+        "41019912345"
+      ],
+      "falsePositiveNotes": "Very broad match — 9 digits",
+      "validate": "dNummer"
+    },
+    {
+      "id": "regex:no:orgnr",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{9}\\s?(?:MVA)?\\b",
+      "flags": "g",
+      "confidence": 0.6,
+      "domains": [
+        "financial",
+        "legal"
+      ],
+      "description": "Norwegian organization number (9 digits, optionally followed by MVA)",
+      "examples": [
+        "123456789",
+        "123456789 MVA"
+      ]
+    },
+    {
+      "id": "regex:no:phone",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?47[\\s-]?)?\\d{2}[\\s-]?\\d{2}[\\s-]?\\d{2}[\\s-]?\\d{2}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Norwegian phone number (8 digits, optionally with +47)",
+      "examples": [
+        "+47 12 34 56 78",
+        "12345678"
+      ]
+    },
+    {
+      "id": "regex:no:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{4}\\s+[\\p{L}][\\p{L}\\s-]+\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Norwegian postal code (4 digits) followed by city name",
+      "examples": [
+        "0001 Oslo",
+        "5003 Bergen"
+      ]
+    },
+    {
+      "id": "regex:no:currency_words",
+      "entityType": "CURRENCY",
+      "pattern": "(?:(?:med\\s+)?(?:bokstaver|ord)\\s*:\\s*|(?:summen|beløpet)\\s+)[\\p{L}\\s-]+(?:kroner|øre)\\b",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Norwegian amount written in words (e.g., \"med bokstaver: åtte tusen fem hundre kroner\")",
+      "examples": [
+        "med bokstaver: åtte tusen fem hundre kroner"
+      ]
+    },
+    {
+      "id": "regex:no:street",
+      "entityType": "ADDRESS",
+      "pattern": "[\\p{L}][\\p{L}\\s-]*(?:gata|gaten|gate|veien|vegen|vei|veg|stien|torget|plassen|plass|allé|alleen)\\s+\\d+[\\p{L}]?",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "contact"
+      ],
+      "description": "Norwegian street address (name + gate/veien/vei/etc. + number)",
+      "examples": [
+        "Karl Johans gate 22",
+        "Storgata 15",
+        "Bygdøy allé 3"
+      ]
+    },
+    {
+      "id": "regex:no:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s&.']+(?:\\s+)?(?:AS|ASA|ANS|DA)\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Norwegian company name with legal form (AS, ASA, ANS, DA)",
+      "examples": [
+        "Equinor ASA",
+        "Telenor ASA",
+        "Nordea Bank ASA"
+      ]
+    }
+  ]
+}

package/rules/pl.json

@@ -0,0 +1,210 @@
+{
+  "region": "pl",
+  "rules": [
+    {
+      "id": "regex:pl:pesel",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{2}[0-3]\\d[0-3]\\d{6}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "identity",
+        "hr",
+        "medical"
+      ],
+      "description": "Polish PESEL national identification number (11 digits, YYMMDD + 5)",
+      "examples": [
+        "89052310002",
+        "02271400004"
+      ],
+      "validate": "pesel"
+    },
+    {
+      "id": "regex:pl:id_card",
+      "entityType": "SSN",
+      "pattern": "\\b[A-Z]{3}\\s?\\d{6}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "identity"
+      ],
+      "description": "Polish ID card number (3 letters + 6 digits)",
+      "examples": [
+        "ABC 523456"
+      ],
+      "validate": "plIdCard"
+    },
+    {
+      "id": "regex:pl:passport",
+      "entityType": "SSN",
+      "pattern": "\\b[A-Z]{2}\\s?\\d{7}\\b",
+      "flags": "g",
+      "confidence": 0.7,
+      "domains": [
+        "identity"
+      ],
+      "description": "Polish passport number (2 letters + 7 digits)",
+      "examples": [
+        "AB 1234567",
+        "CD1234567"
+      ]
+    },
+    {
+      "id": "regex:pl:nip",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{3}-?\\d{3}-?\\d{2}-?\\d{2}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "financial",
+        "legal"
+      ],
+      "description": "Polish NIP tax identification number (10 digits, XXX-XXX-XX-XX)",
+      "examples": [
+        "123-456-78-19",
+        "1234567819"
+      ],
+      "validate": "nip"
+    },
+    {
+      "id": "regex:pl:regon",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{9}(?:\\d{5})?\\b",
+      "flags": "g",
+      "confidence": 0.6,
+      "domains": [
+        "financial",
+        "legal"
+      ],
+      "description": "Polish REGON business registry number (9 or 14 digits)",
+      "examples": [
+        "123456785"
+      ],
+      "falsePositiveNotes": "9-digit REGON overlaps with many other number formats",
+      "validate": "regon"
+    },
+    {
+      "id": "regex:pl:bank_account",
+      "entityType": "OTHER",
+      "pattern": "\\b\\d{4}\\s?\\d{4}\\s?\\d{4}\\s?\\d{4}\\s?\\d{4}\\s?\\d{4}\\s?\\d{2}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "financial"
+      ],
+      "description": "Polish bank account number (26 digits, often spaced in groups of 4)",
+      "examples": [
+        "1234 5678 9012 3456 7890 1234 56"
+      ]
+    },
+    {
+      "id": "regex:pl:phone",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?48[\\s-]?)?\\d{3}[\\s-]?\\d{3}[\\s-]?\\d{3}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Polish phone number (9 digits, optionally with +48 prefix)",
+      "examples": [
+        "+48 600 123 456",
+        "600-123-456",
+        "48 600123456"
+      ]
+    },
+    {
+      "id": "regex:pl:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{2}-\\d{3}\\s+[\\p{L}][\\p{L}\\s-]+\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Polish postal code (XX-XXX) followed by city name",
+      "examples": [
+        "00-950 Warszawa",
+        "31-501 Kraków",
+        "80-244 Gdańsk Wrzeszcz"
+      ]
+    },
+    {
+      "id": "regex:pl:currency_words",
+      "entityType": "CURRENCY",
+      "pattern": "(?:słownie:\\s*)[\\p{L}\\s]+(?:złot(?:ych|y|e)|groszy|grosz(?:e)?)\\b",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Polish amount written in words after \"słownie:\" (e.g., \"słownie: osiem tysięcy pięćset złotych\")",
+      "examples": [
+        "słownie: osiem tysięcy pięćset złotych"
+      ]
+    },
+    {
+      "id": "regex:pl:street",
+      "entityType": "ADDRESS",
+      "pattern": "(?:ul\\.|al\\.|pl\\.|os\\.)\\s+[\\p{L}][\\p{L}\\s]+\\s+\\d+[\\p{L}]?(?:\\s*[/\\\\]\\s*\\d+[\\p{L}]?)?",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "contact"
+      ],
+      "description": "Polish street address (ul./al./pl./os. + name + number)",
+      "examples": [
+        "ul. Floriańska 27/3",
+        "ul. Juliusza Słowackiego 15/8",
+        "al. Krakowska 42",
+        "os. Złotego Wieku 12"
+      ]
+    },
+    {
+      "id": "regex:pl:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s]+(?:Sp(?:ółka)?\\s*z\\s*o\\.?\\s*o\\.?|Spółka\\s+z\\s+ograniczon[\\p{L}]+\\s+odpowiedzialno[\\p{L}]+|Sp\\.\\s*j\\.|S\\.A\\.|Sp\\.\\s*k\\.)",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Polish company name with legal form (Sp. z o.o., S.A., Sp. j., Sp. k.)",
+      "examples": [
+        "Santander Bank Polska S.A.",
+        "NOVAMED Spółka z ograniczoną odpowiedzialnością"
+      ]
+    },
+    {
+      "id": "regex:pl:court_case",
+      "entityType": "OTHER",
+      "pattern": "\\b[IVXLCDM]+\\s+(?:K|C|Ca|Cz|Co|Gz|Ga|GCo|GC)\\s+\\d+\\/\\d{2,4}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "legal"
+      ],
+      "description": "Polish court case signature (e.g., \"III K 123/24\", \"I C 456/2023\")",
+      "examples": [
+        "III K 123/24",
+        "I C 456/2023"
+      ]
+    },
+    {
+      "id": "regex:pl:krs",
+      "entityType": "OTHER",
+      "pattern": "\\b\\d{4}\\/\\d{8}\\/\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Polish KRS company registry number",
+      "examples": [
+        "0000/12345678/0001"
+      ]
+    }
+  ]
+}