Predefined Entities are Netskope-maintained entities covering common regulatory and sensitive data types, such as payment card numbers, national ID numbers, and financial account formats. Netskope now also provides Predefined AI Entities, which use Named Entity Recognition (NER) to detect unstructured, contextual data that pattern matching alone can’t reliably catch.
Predefined Entities
Predefined Entities are organized into categories with descriptive names and tags for easy search. For example, the internal address-us identifier is surfaced as Postal Addresses (US) and grouped under Postal Addresses.
Key characteristics:
- Multiple identifiers can be combined using AND and OR operators within a single rule.
- You can’t edit a Predefined Data Identifier directly, but you can reference one inside a Custom Data Identifier using {{predefined_data_identifier}} syntax. For more information, see Select Custom Entities.
- A complete, current list of names, categories, and definitions is maintained on the customer-only support site, DLP Predefined Identifiers.
To select predefined entities in a rule:
- In the Entities page of the New DLP Rule workflow, expand a category to view its identifiers. For example, the Financial Data category lists rules for payment cards, including credit card numbers with 12 to 19 digits and different delimiter types.
- Click an identifier to view its detection details.
- Select the identifiers you need, then click Next.
Predefined AI Entities (Named Entity Recognition)
Pattern matching works well for structured data with a fixed shape, such as a 16-digit card number. It doesn’t work as well for unstructured, contextual data, such as a person’s name inside a paragraph or a genomic identifier inside a lab report. Predefined AI Entities close this gap using Named Entity Recognition (NER), a language model that identifies entities by context instead of a fixed pattern.
How NER-Based Entities Differ from Pattern-Based Entities
| Pattern-Based Entities | Predefined AI Entities (NER) | |
|---|---|---|
| Detection method | Regex, keyword dictionary, checksum validators | Contextual language model classification |
| Best for | Fixed-format data, such as IDs and card numbers | Free-text, variable-format data, such as names and clinical terminology |
| Editable | Custom entities are editable; predefined entities can be referenced but not edited | Not editable; predefined AI entities are used as-is |
| Typical grouping | Category, such as Financial Data | Domain, such as PII or Genomic |
Supported Domains
Predefined AI Entities are grouped by detection domain rather than by data format.
| Domain | Example Detections |
|---|---|
| PII | Person names, physical addresses, and other free-text personal identifiers |
| Genomic | Genomic sequence and variant identifiers and related terminology |
| Human-omic | Broader human-omics data, such as proteomic terminology in unstructured records |
Within each domain, entities come in two scopes:
- Un-bundled entities cover a single term class or jurisdiction. Use these when a rule needs a tightly scoped detector, to minimize false positives.
- Bundled entities cover every variant within a domain, such as all genomic identifiers. Use these as a broad backstop in a rule.
Using Predefined AI Entities in a Rule
- In the Entities page of the New DLP Rule workflow, locate the AI Entities category alongside the predefined entities.
- Expand a domain, such as PII or Genomic, to view the available entities.
- Select the entities that fit your use case. Combine an un-bundled entity for precision or a bundled entity for broad coverage.
- Click Next.
You can combine a Predefined AI Entity with a pattern-based custom entity in the same rule to raise confidence before a rule triggers a violation. For example, pair a genomic identifier entity with a custom dictionary of lab codenames using a NEAR condition.