Faasera Documentation

Faasera Synthetic Data Generation — User Guide

This guide explains how to use the Faasera Synthetic Data module to generate safe, realistic, and lineage-aware test data for development, analytics, and QA use cases.

---

Overview

Faasera Synthetic Data Generator allows users to:

• Replace sensitive data with fake but realistic values.
• Preserve referential integrity across related tables.
• Generate synthetic data that mimics original statistical distribution.
• Use the same policy engine as masking, ensuring consistent treatment.

---

Key Features

Feature	Description
Seed-Based Generation	Uses seed files for deterministic or constrained value sets.
Field-Aware Logic	Applies domain-specific generation (e.g., names, dates, credit cards).
Lineage Consistency	Ensures synthetic data is relationally consistent across joins.
Custom Rule Support	Define custom generators for business-specific entities.

---

Supported Generator Types

Generator Type	Use Case Example	Notes
FAKE_NAME	First/Last names	Locale-aware
FAKE_EMAIL	Email addresses	Optional domain control
FAKE_DATE	Birthdates, registration dates	With configurable range
FAKE_CREDIT_CARD	Dummy card numbers (Luhn-valid)	Brand-specific supported
SEED_BASED	From seed file with column constraints	Deterministic or random from list
CUSTOM_LOGIC	Regex or lookup-based	Requires plugin or lambda function support

---

Configuration Options

Synthetic data generation is governed by the masking policy, using maskFunction.type = GENERATE.

Example Policy Snippet

{
  "columnRules": {
    "email": {
      "maskFunction": {
        "type": "FAKE_EMAIL"
      }
    },
    "customer_type": {
      "maskFunction": {
        "type": "SEED_BASED",
        "seedFileName": "customer_type.csv",
        "seedSeparator": ",",
        "seedFieldPosition": 0
      }
    }
  }
}

Runtime Hints (Optional)

You can control generation using:

• Locale: Affects names, cities, dates.
• Null Probability: Simulates real-world missing values.
• Format Templates: For phone, date, or ID number styles.

---

Integration Modes

Mode	Description
In-place	Replace original column values in source DB
Source → Target	Populate a new database or table with synthetic data
Standalone	Generate data samples without an original dataset

---

Usage Scenarios

Scenario	Synthetic Approach
Load testing with fake users	FAKE_NAME + FAKE_EMAIL
Sensitive system without masking	GENERATE mode for all PII
Replace all values with lineage	SEED_BASED + Referential Map
Generate GDPR/CCPA-compliant datasets	CUSTOM_LOGIC + audit trail

---

Sample Output

Input Row:

{
  "name": "John Smith",
  "email": "john@real.com",
  "dob": "1980-01-01"
}

After Generation:

{
  "name": "Alice Williams",
  "email": "alice@testmail.io",
  "dob": "1994-09-23"
}

---

Compliance Notes

• All synthetic values are non-reversible unless seed-based determinism is explicitly configured.
• Generation rules can be exported and versioned for audit purposes.
• Referential joins are maintained by maskKeyColumn + compositeKeyColumn mapping.

---

Best Practices

• Use SEED_BASED generation to ensure test consistency.
• Avoid overfitting generated data to real records.
• Add nullProbability for realism.
• Combine masking and generation in hybrid environments.

---

For advanced setup or consulting, contact the Faasera team at www.faasera.ai