Faasera Documentation

🧪 Faasera Profiler Guide

Faasera Profiler supports advanced techniques to detect, classify, and protect sensitive data in structured and unstructured text. You can configure the Profiler to use AI-driven entity hinting, heuristic recognition, or a hybrid of both approaches.

This guide explains the available modes, how to configure them, and how Faasera blends techniques like dictionary matching, NLP models, and checksum validation to enrich entity recognition.

Hint Processing Modes

Mode Description
AI_ONLY Use only AI hint service results. Disables all heuristic detection.
HEURISTIC_ONLY Use only heuristic recognizers (dictionary, NLP, checksum, etc.).
HYBRID (default) Combine both AI and heuristic results, merging them with disambiguation.

AI Hinting (FaaseraAI Hint Service)

When enabled, Faasera sends selected text values to an external Hinting Service (e.g., GLiNER) via HTTP POST. The service returns structured entity spans like this:

[
  {
    "start": 0,
    "end": 5,
    "text": "John",
    "label": "first_name",
    "score": 0.7638
  }
]

These spans are mapped to internal recognizer types:

"classificationToHintService": {
  "FIRST_NAME": "first_name",
  "EMAIL_ADDRESS": "email",
  "SSN": "ssn"
}

Heuristic Recognition Techniques

1. NLP Models (Apache OpenNLP)

Entity Type Example
Person Name “My name is Alice Carter” → Alice Carter
Location “I live in New York” → New York

2. Dictionary-Based Matching

Custom domain dictionaries:

{
  "COUNTRY_NAMES": ["Australia", "Canada"],
  "HEALTH_TERMS": ["asthma", "diabetes"]
}

Input: “The patient was diagnosed with diabetes in Australia.”
Output: diabetes → Health Term, Australia → Country Name

3. Keyword-Based Recognition

Trigger context keywords like:

4. Checksum-Based Validation

Type Example Method
Credit Card 4111 1111 1111 1111 Luhn Algorithm
TFN 123 456 782 Modulus check

5. Regex-Based Matching (Optional)

Define advanced expressions like:

[A-Z]{3}-\d{4}

⚙️ Why Use Heuristics?

Benefit Description
Speed No external API calls
Predictability Same input → same result
Auditability Easy to trace
Customizability User-defined logic
Fallback Works even if AI model fails

Detection Workflow

               Raw Text Input
                      ↓
              Faasera Profiler
        ┌──────────┬───────────┬───────────┐
        │  AI Hint │ Heuristics│ Config    │
        └────┬─────┴────┬──────┴────┬──────┘
             ↓          ↓           ↓
     Hint Spans  Heuristic Spans  Rules
             ↓          ↓           ↓
        Span Disambiguation
             ↓
   Filter → Sort → Merge Confidence
             ↓
       Final Spans + Redaction

Configuration

In Policy JSON

"aiModel": {
  "enabled": true,
  "hintProcessingMode": "HYBRID",
  "classificationToHintService": {
    "FIRST_NAME": "first_name",
    "EMAIL_ADDRESS": "email",
    "SSN": "ssn"
  }
}

Confidence & Span Handling

Use Case Scenarios

Use Case Recommended Mode
AI-trained model use AI_ONLY
Regulated pipelines HEURISTIC_ONLY
General-purpose HYBRID

Output Example

Input:

"John Doe’s email is john@faasera.ai and SSN is 123-45-6789."

Output:

[
  { "text": "John", "label": "FIRST_NAME", "confidence": 0.76 },
  { "text": "john@faasera.ai", "label": "EMAIL_ADDRESS", "confidence": 0.99 },
  { "text": "123-45-6789", "label": "SSN", "confidence": 0.95 }
]

Summary of Components

Component Purpose
HintService Sends text to AI
HintMapper Maps labels to internal recognizers
Profiler Combines and redacts spans

Final Notes