@doccloak/core 0.8.0

package/rules/pt.json

@@ -0,0 +1,113 @@
+{
+  "region": "pt",
+  "rules": [
+    {
+      "id": "regex:pt:cc",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{9}\\s?[A-Z]{2}\\d\\b",
+      "flags": "gi",
+      "confidence": 0.85,
+      "domains": [
+        "identity"
+      ],
+      "description": "Portuguese Cartão de Cidadão number (citizen card)",
+      "examples": [
+        "123456789 ZZ1"
+      ]
+    },
+    {
+      "id": "regex:pt:nif",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{9}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "financial",
+        "identity"
+      ],
+      "description": "Portuguese NIF tax number (9 digits)",
+      "examples": [
+        "123456789"
+      ],
+      "falsePositiveNotes": "9 digits is very broad",
+      "validate": "nif"
+    },
+    {
+      "id": "regex:pt:phone",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?351[\\s-]?)?\\d{3}[\\s-]?\\d{3}[\\s-]?\\d{3}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Portuguese phone number (9 digits, optionally with +351)",
+      "examples": [
+        "+351 912 345 678",
+        "912-345-678"
+      ]
+    },
+    {
+      "id": "regex:pt:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{4}-\\d{3}\\s+[\\p{L}][\\p{L}\\s'-]+\\b",
+      "flags": "gu",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "Portuguese postal code (XXXX-XXX) followed by city name",
+      "examples": [
+        "1000-001 Lisboa",
+        "4000-322 Porto"
+      ]
+    },
+    {
+      "id": "regex:pt:currency_words",
+      "entityType": "CURRENCY",
+      "pattern": "(?:(?:por\\s+)?extenso\\s*:\\s*|(?:a\\s+)?(?:quantia|importância|soma)\\s+de\\s+)[\\p{L}\\s-]+(?:euros?|cêntimos?|reais|centavos?)\\b",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Portuguese amount written in words (e.g., \"por extenso: oito mil e quinhentos euros\")",
+      "examples": [
+        "por extenso: oito mil e quinhentos euros"
+      ]
+    },
+    {
+      "id": "regex:pt:street",
+      "entityType": "ADDRESS",
+      "pattern": "(?:(?:Rua|R\\.|Avenida|Av\\.|Praça|Pç\\.|Travessa|Tv\\.|Largo|Lg\\.|Alameda|Al\\.|Estrada|Est\\.)\\s+)[\\p{L}][\\p{L}\\s'-]+(?:,?\\s*(?:n\\.?º?\\s*)?\\d+)?",
+      "flags": "giu",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "Portuguese street address (Rua/Avenida/Praça + name + optional number)",
+      "examples": [
+        "Rua Augusta 42",
+        "Av. da Liberdade 10",
+        "Praça do Comércio"
+      ]
+    },
+    {
+      "id": "regex:pt:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s&.']+(?:\\s+)?(?:Lda\\.?|S\\.A\\.?|SGPS)\\b",
+      "flags": "gu",
+      "confidence": 0.9,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Portuguese company name with legal form (Lda., S.A., SGPS)",
+      "examples": [
+        "Galp Energia S.A.",
+        "Construções Silva Lda.",
+        "EDP SGPS"
+      ]
+    }
+  ]
+}

package/rules/schema.json

@@ -0,0 +1,91 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://github.com/WLojek/DocCloak.Core/rules/schema.json",
+  "title": "DocCloak shared regex rule pack",
+  "description": "Canonical cross-runtime rule format (architecture doc section 6). One file per region plus universal.json. Patterns are restricted to the ECMAScript/Python common subset; flags live in the flags field (no inline flags); validate is a name resolved against the per-runtime validator registry (validators.ts / validators.py).",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["region", "rules"],
+  "properties": {
+    "region": {
+      "type": "string",
+      "pattern": "^(universal|[a-z]{2})$"
+    },
+    "rules": {
+      "type": "array",
+      "minItems": 1,
+      "items": { "$ref": "#/$defs/rule" }
+    }
+  },
+  "$defs": {
+    "rule": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["id", "entityType", "pattern", "flags", "confidence", "domains", "description"],
+      "properties": {
+        "id": {
+          "description": "Stable detector identifier, e.g. regex:pl:pesel",
+          "type": "string",
+          "pattern": "^regex:(universal|[a-z]{2}):[a-z0-9_-]+$"
+        },
+        "entityType": {
+          "enum": [
+            "PERSON",
+            "EMAIL",
+            "PHONE",
+            "SSN",
+            "CREDIT_CARD",
+            "DATE",
+            "CURRENCY",
+            "IP_ADDRESS",
+            "IBAN",
+            "ADDRESS",
+            "COMPANY",
+            "OTHER"
+          ]
+        },
+        "pattern": {
+          "description": "Regex source in the ECMAScript/Python common subset (compiled per runtime)",
+          "type": "string",
+          "minLength": 1
+        },
+        "flags": {
+          "description": "Regex flags; the g flag is mandatory, inline flags in the pattern are forbidden",
+          "type": "string",
+          "pattern": "^g[imsuy]*$"
+        },
+        "confidence": {
+          "type": "number",
+          "exclusiveMinimum": 0,
+          "maximum": 1
+        },
+        "domains": {
+          "type": "array",
+          "minItems": 1,
+          "items": {
+            "enum": ["financial", "medical", "legal", "hr", "identity", "contact", "technical", "general"]
+          }
+        },
+        "description": {
+          "type": "string",
+          "minLength": 1
+        },
+        "examples": {
+          "description": "Strings that MUST match the compiled pattern (and pass validate, if any); drives the cross-runtime conformance test",
+          "type": "array",
+          "minItems": 1,
+          "items": { "type": "string", "minLength": 1 }
+        },
+        "falsePositiveNotes": {
+          "type": "string",
+          "minLength": 1
+        },
+        "validate": {
+          "description": "Name of a post-match validation function in the runtime validator registry",
+          "type": "string",
+          "pattern": "^[a-z][a-zA-Z0-9]*$"
+        }
+      }
+    }
+  }
+}

package/rules/se.json

@@ -0,0 +1,149 @@
+{
+  "region": "se",
+  "rules": [
+    {
+      "id": "regex:se:personnummer",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{6}[-+]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "identity",
+        "hr",
+        "medical"
+      ],
+      "description": "Swedish personnummer (YYMMDD-XXXX or YYMMDDXXXX)",
+      "examples": [
+        "850523-0006",
+        "8505230006"
+      ],
+      "validate": "personnummer"
+    },
+    {
+      "id": "regex:se:personnummer_12",
+      "entityType": "SSN",
+      "pattern": "\\b(?:19|20)\\d{6}[-+]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "identity",
+        "hr",
+        "medical"
+      ],
+      "description": "Swedish personnummer 12-digit format (YYYYMMDD-XXXX)",
+      "examples": [
+        "19850523-1478",
+        "198505231478"
+      ]
+    },
+    {
+      "id": "regex:se:samordningsnummer",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{6}[-+]?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.7,
+      "domains": [
+        "identity"
+      ],
+      "description": "Swedish samordningsnummer (coordination number, day +60)",
+      "examples": [
+        "850583-1478"
+      ],
+      "falsePositiveNotes": "Same format as personnummer — differentiated by day > 60",
+      "validate": "samordningsnummer"
+    },
+    {
+      "id": "regex:se:organisationsnummer",
+      "entityType": "SSN",
+      "pattern": "\\b\\d{6}-?\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.6,
+      "domains": [
+        "financial",
+        "legal"
+      ],
+      "description": "Swedish organisationsnummer (company number, 10 digits)",
+      "examples": [
+        "556036-0793"
+      ],
+      "falsePositiveNotes": "Same 10-digit format as personnummer"
+    },
+    {
+      "id": "regex:se:phone",
+      "entityType": "PHONE",
+      "pattern": "\\b(?:\\+?46[\\s-]?)?0?\\d{1,3}[\\s-]?\\d{2,3}[\\s-]?\\d{2}[\\s-]?\\d{2}\\b",
+      "flags": "g",
+      "confidence": 0.75,
+      "domains": [
+        "contact"
+      ],
+      "description": "Swedish phone number (variable-length area code)",
+      "examples": [
+        "+46 8 123 45 67",
+        "070-123 45 67"
+      ]
+    },
+    {
+      "id": "regex:se:postal",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{3}\\s?\\d{2}\\s+[\\p{L}][\\p{L}\\s-]+\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "contact"
+      ],
+      "description": "Swedish postal code (XXX XX) followed by city name",
+      "examples": [
+        "111 22 Stockholm",
+        "41301 Göteborg"
+      ]
+    },
+    {
+      "id": "regex:se:currency_words",
+      "entityType": "CURRENCY",
+      "pattern": "(?:(?:med\\s+)?(?:bokstäver|ord)\\s*:\\s*|(?:summan|beloppet)\\s+)[\\p{L}\\s-]+(?:kronor|öre)\\b",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Swedish amount written in words (e.g., \"med bokstäver: åttatusen femhundra kronor\")",
+      "examples": [
+        "med bokstäver: åttatusen femhundra kronor"
+      ]
+    },
+    {
+      "id": "regex:se:street",
+      "entityType": "ADDRESS",
+      "pattern": "[\\p{L}][\\p{L}\\s-]*(?:gatan|vägen|stigen|torget|platsen|gränd|backen|liden|ängen)\\s+\\d+[\\p{L}]?",
+      "flags": "giu",
+      "confidence": 0.9,
+      "domains": [
+        "contact"
+      ],
+      "description": "Swedish street address (name + gatan/vägen/etc. + number)",
+      "examples": [
+        "Drottninggatan 42",
+        "Sveavägen 15",
+        "Stortorget 3"
+      ]
+    },
+    {
+      "id": "regex:se:company",
+      "entityType": "COMPANY",
+      "pattern": "[\\p{L}][\\p{L}\\s&.']+(?:\\s+)?(?:AB|HB|KB)\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "legal",
+        "financial"
+      ],
+      "description": "Swedish company name with legal form (AB, HB, KB)",
+      "examples": [
+        "Volvo AB",
+        "Handelsbanken AB",
+        "Ericsson AB"
+      ]
+    }
+  ]
+}

package/rules/universal.json

@@ -0,0 +1,279 @@
+{
+  "region": "universal",
+  "rules": [
+    {
+      "id": "regex:universal:email",
+      "entityType": "EMAIL",
+      "pattern": "[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}",
+      "flags": "g",
+      "confidence": 0.95,
+      "domains": [
+        "contact"
+      ],
+      "description": "Email address",
+      "examples": [
+        "john@example.com",
+        "user.name+tag@domain.co.uk"
+      ]
+    },
+    {
+      "id": "regex:universal:phone_intl",
+      "entityType": "PHONE",
+      "pattern": "\\+\\d[\\d\\s()-]{7,18}\\d",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "contact"
+      ],
+      "description": "International phone number with + prefix",
+      "examples": [
+        "+1 (555) 123-4567",
+        "+48 600 123 456",
+        "+44 20 7946 0958"
+      ]
+    },
+    {
+      "id": "regex:universal:iban",
+      "entityType": "IBAN",
+      "pattern": "\\b[A-Z]{2}\\s?\\d{2}[\\s]?[A-Z\\d\\s]{10,30}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "IBAN (2-letter country code + 2 check digits + up to 30 alphanumeric)",
+      "examples": [
+        "GB29 NWBK 6016 1331 9268 19",
+        "DE89 3704 0044 0532 0130 00",
+        "PL61 1090 1014 0000 0712 1981 2874"
+      ],
+      "validate": "ibanMod97"
+    },
+    {
+      "id": "regex:universal:credit_card",
+      "entityType": "CREDIT_CARD",
+      "pattern": "\\b\\d{4}[\\s-]?\\d{4}[\\s-]?\\d{4}[\\s-]?\\d{1,7}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "financial"
+      ],
+      "description": "Credit/debit card number (13-19 digits, possibly spaced/dashed)",
+      "examples": [
+        "4532 0151 1283 0366",
+        "5425-2334-3010-9903"
+      ],
+      "falsePositiveNotes": "Luhn validation reduces false positives from random digit sequences",
+      "validate": "luhn"
+    },
+    {
+      "id": "regex:universal:currency_symbol_prefix",
+      "entityType": "CURRENCY",
+      "pattern": "[$€£¥₹₽₩₺₴]\\s?\\d{1,3}(?:[,.\\s]\\d{3})*(?:[.,]\\d{1,2})?\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "financial"
+      ],
+      "description": "Currency amount with leading symbol ($, €, £, ¥, ₹, ₽, ₩, ₺, ₴)",
+      "examples": [
+        "$45,000",
+        "€1.234,56",
+        "£100.00",
+        "¥50,000"
+      ]
+    },
+    {
+      "id": "regex:universal:currency_code_suffix",
+      "entityType": "CURRENCY",
+      "pattern": "\\b\\d{1,3}(?:[,.\\s]\\d{3})*(?:[.,]\\d{1,2})?\\s?(?:USD|EUR|GBP|CHF|JPY|PLN|SEK|NOK|DKK|CZK|HUF|RON|BGN|HRK|RUB|UAH|TRY|BRL|ARS|MXN|COP|CLP|PEN|INR|CNY|KRW|AUD|CAD|NZD|ZAR|SGD|HKD|TWD|THB|MYR|IDR|PHP|VND|AED|SAR|QAR|KWD|BHD|OMR|ILS|EGP|NGN|KES|GHS|TZS|UGX)\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "financial"
+      ],
+      "description": "Currency amount with trailing ISO code (e.g., \"1,000.00 USD\")",
+      "examples": [
+        "1,000.00 USD",
+        "45 000 PLN",
+        "100.50 EUR"
+      ]
+    },
+    {
+      "id": "regex:universal:currency_symbol_suffix",
+      "entityType": "CURRENCY",
+      "pattern": "\\b\\d{1,3}(?:[,.\\s]\\d{3})*(?:[.,]\\d{1,2})?\\s?(?:zł|kr|Kč|lei|лв|Ft|kn|₽|грн|₺|R\\$|S\\/\\.|руб|₹|元|圆|円|원|ر\\.س|ر\\.ق|د\\.إ|₪|₦|₵|₱|₫|₸|₼|₾|฿|RM|Rp|đ)(?=\\s|$|[.,;:!?)}\\]])",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "financial"
+      ],
+      "description": "Currency amount with trailing local symbol (zł, kr, Kč, lei, etc.)",
+      "examples": [
+        "8 500,00 zł",
+        "1 200 kr",
+        "3.500 Kč",
+        "10 000 lei",
+        "5 000 Ft"
+      ]
+    },
+    {
+      "id": "regex:universal:ipv4",
+      "entityType": "IP_ADDRESS",
+      "pattern": "\\b(?:(?:25[0-5]|2[0-4]\\d|[01]?\\d\\d?)\\.){3}(?:25[0-5]|2[0-4]\\d|[01]?\\d\\d?)\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "technical"
+      ],
+      "description": "IPv4 address",
+      "examples": [
+        "192.168.1.1",
+        "10.0.0.255",
+        "172.16.0.1"
+      ],
+      "validate": "ipOctets"
+    },
+    {
+      "id": "regex:universal:ipv6",
+      "entityType": "IP_ADDRESS",
+      "pattern": "\\b(?:[0-9a-fA-F]{1,4}:){7}[0-9a-fA-F]{1,4}\\b",
+      "flags": "g",
+      "confidence": 0.9,
+      "domains": [
+        "technical"
+      ],
+      "description": "IPv6 address (full form)",
+      "examples": [
+        "2001:0db8:85a3:0000:0000:8a2e:0370:7334"
+      ]
+    },
+    {
+      "id": "regex:universal:mac_address",
+      "entityType": "OTHER",
+      "pattern": "\\b[0-9a-fA-F]{2}(?::[0-9a-fA-F]{2}){5}\\b",
+      "flags": "g",
+      "confidence": 0.85,
+      "domains": [
+        "technical"
+      ],
+      "description": "MAC address (colon-separated)",
+      "examples": [
+        "00:1B:44:11:3A:B7"
+      ]
+    },
+    {
+      "id": "regex:universal:date_numeric",
+      "entityType": "DATE",
+      "pattern": "\\b\\d{1,2}[./]\\d{1,2}[./]\\d{4}\\b",
+      "flags": "g",
+      "confidence": 0.75,
+      "domains": [
+        "general"
+      ],
+      "description": "Numeric date (dd.mm.yyyy, dd/mm/yyyy, mm/dd/yyyy)",
+      "examples": [
+        "15.03.2025",
+        "03/15/2025",
+        "1.1.2024"
+      ],
+      "falsePositiveNotes": "Cannot distinguish dd/mm/yyyy from mm/dd/yyyy — both valid"
+    },
+    {
+      "id": "regex:universal:date_iso",
+      "entityType": "DATE",
+      "pattern": "\\b\\d{4}-\\d{2}-\\d{2}\\b",
+      "flags": "g",
+      "confidence": 0.8,
+      "domains": [
+        "general"
+      ],
+      "description": "ISO 8601 date (yyyy-mm-dd)",
+      "examples": [
+        "2025-03-15",
+        "1990-01-01"
+      ]
+    },
+    {
+      "id": "regex:universal:date_word",
+      "entityType": "DATE",
+      "pattern": "\\b\\d{1,2}\\s+[\\p{L}]{3,}\\s+\\d{4}\\b",
+      "flags": "gu",
+      "confidence": 0.8,
+      "domains": [
+        "general"
+      ],
+      "description": "Date with month name in any language (e.g., \"10 marca 2025\", \"15 March 2025\")",
+      "examples": [
+        "10 March 2025",
+        "15 marca 2025",
+        "1 janvier 2024"
+      ]
+    },
+    {
+      "id": "regex:universal:date_month_first",
+      "entityType": "DATE",
+      "pattern": "\\b(?:Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|Jun(?:e)?|Jul(?:y)?|Aug(?:ust)?|Sep(?:tember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\\s+\\d{1,2},?\\s+\\d{4}\\b",
+      "flags": "gi",
+      "confidence": 0.85,
+      "domains": [
+        "general"
+      ],
+      "description": "English date with month name first (e.g., \"March 15, 2025\")",
+      "examples": [
+        "March 15, 2025",
+        "Jan 1 2024",
+        "December 25, 2023"
+      ]
+    },
+    {
+      "id": "regex:universal:labeled_id",
+      "entityType": "SSN",
+      "pattern": "(?<=[\\p{L}][\\p{L}\\s]{0,30}[:#]\\s?)[A-Z0-9]{2,4}[\\s-]?[\\d]{4,}(?:[\\s-][\\d]+)*",
+      "flags": "giu",
+      "confidence": 0.7,
+      "domains": [
+        "identity",
+        "general"
+      ],
+      "description": "Labeled identifier (Key: Value pattern, e.g., \"ID: ABC12345\", \"Ref: 987654\")",
+      "examples": [
+        "ID: ABC12345"
+      ],
+      "falsePositiveNotes": "Broad pattern — catches many key:value pairs that may not be PII"
+    },
+    {
+      "id": "regex:universal:long_number",
+      "entityType": "SSN",
+      "pattern": "(?<![.\\d])\\b\\d[\\d\\s-]{7,}\\d\\b(?![.\\d])",
+      "flags": "g",
+      "confidence": 0.5,
+      "domains": [
+        "identity",
+        "financial"
+      ],
+      "description": "Standalone long digit sequence (8+ digits) — likely IDs, account numbers, case numbers",
+      "examples": [
+        "123456789",
+        "123-456-789-00"
+      ],
+      "falsePositiveNotes": "Low confidence — many non-PII numbers match. Overlap resolution with specific patterns prevents double-detection."
+    },
+    {
+      "id": "regex:universal:postal_city",
+      "entityType": "ADDRESS",
+      "pattern": "\\b\\d{2,5}[-\\s]?\\d{2,4}\\s+[\\p{L}][\\p{L}]+\\b",
+      "flags": "gu",
+      "confidence": 0.65,
+      "domains": [
+        "contact"
+      ],
+      "description": "Postal/zip code followed by city name (various country formats)",
+      "examples": [
+        "00-950 Warszawa",
+        "75008 Paris"
+      ],
+      "falsePositiveNotes": "May match non-address number+word sequences"
+    }
+  ]
+}