Auto-Detect and Mask
The default masking mode for free-form text — Protecto scans input, detects sensitive entities by policy, and returns a masked version with entity tags.
Auto-Detect and Mask is the most commonly used mode in Protecto. It is designed for situations where sensitive data appears in free-form text and you don't want to pre-classify fields.
This mode combines detection and masking into a single operation.
What it does
Auto-Detect and Mask performs four steps automatically:
- Scans the input text
- Detects sensitive values based on policy rules
- Classifies each detected value by entity type
- Replaces detected values with tokens wrapped in entity tags
You do not need to specify entity types, token names, or formats. All behavior is policy-driven.
Example
Input:
User John Doe (john.doe@example.com) requested a refund on 15/8/2010
Masked output:
User <PERSON>VJYe 03W</PERSON> (<EMAIL>0gN3SkjL@0ffM3CDS</EMAIL>) requested a refund on <DATE>5Fd890</DATE>
Names, emails, and dates were detected automatically. Each value was tokenized and entity tags preserved readability.
When to use it
Auto-Detect and Mask is the right choice when:
- Input is unstructured or semi-structured
- Sensitive data appears inline with normal text
- You don't control the schema
- You want the fastest path to safe data handling
Typical use cases: LLM prompts, application logs, chat messages, user-generated content, support notes.
What controls detection behavior
Auto-Detect and Mask is controlled entirely by policy configuration. Policies define:
- Which entity types are detected
- Which custom PII types are included
- How each entity is tokenized
- Whether toxicity analysis runs
Changing a policy updates behavior without changing client code.
Auto-Detect vs explicit masking
| Scenario | Recommended approach |
|---|---|
| Free-form text | Auto-Detect and Mask |
| Known sensitive fields | Explicit masking with token |
| Structured records | Explicit masking with format |
| Internal identifiers | Custom PII detection |
Auto-Detect optimizes for speed and simplicity, not precision control.
Last updated 3 weeks ago
Built with Documentation.AI