Regex Patterns

The Regex detector matches content against one or more regular expression patterns. Matches are exact and deterministic — the same input always produces the same output, with no model inference cost.

When to use

• Structured codes and identifiers — order numbers, ticket IDs, IBAN / credit-card formats, employee IDs, or any token with a known structure.
• Policy phrases and keywords — contract clauses, prohibited terms, or required disclosures that must appear verbatim.
• High-throughput scanning — regex is the fastest detection method in Classifyre; use it wherever deterministic matching is sufficient.
• Compliance audit trails — exact matches are easy to explain to auditors without model interpretability concerns.
• Combining with other detectors — layer a regex detector with a GLiNER2 or LLM detector to handle structured patterns separately from semantic ones.

How it works

The patterns map defines one or more named rules. Each rule specifies a pattern (a regular expression string) and optional metadata. Classifyre applies every pattern to the scanned asset content and creates a finding for each match.

Matches include the matched span, the pattern name, and the source position. You can map pattern names to severity levels using a severity_map if some patterns represent higher-risk matches than others.

Configuration

Examples

IBAN and order number detection

{
  "type": "REGEX",
  "patterns": {
    "IBAN": {
      "pattern": "\\b[A-Z]{2}\\d{2}[A-Z0-9]{4}\\d{7}([A-Z0-9]?){0,16}\\b",
      "description": "International bank account number"
    },
    "OrderNumber": {
      "pattern": "\\bORD-\\d{6,10}\\b",
      "description": "Internal order reference in ORD-NNNNNN format"
    }
  }
}

Prohibited terms with severity mapping

{
  "type": "REGEX",
  "patterns": {
    "ProhibitedVendor": {
      "pattern": "\\b(VendorAlpha|VendorBeta)\\b",
      "description": "Reference to a vendor on the restricted list"
    },
    "EmbargoCurrency": {
      "pattern": "\\b(XBT|ZEC|XMR)\\b",
      "description": "Reference to a sanctioned or embargoed currency token"
    }
  },
  "validation": {
    "case_sensitive": false
  }
}