Transform production databases into privacy-safe shadow copies while preserving relational integrity.

Unlike simple "faker" scripts, Mimicry uses Deterministic Hashing. If "John Doe" appears in the Users table and the Logs table, he becomes "Alice Smith" in both places — preserving your ability to run joins, analytics, and tests without exposing real identities.

+------------------+                    +------------------+
|  Production DB   |                    |   Shadow DB      |
+------------------+     Mimicry        +------------------+
| john@corp.com    | ----------------->  | alex42@test.com  |
| Jane Smith       |                    | Morgan Davis     |
| +1 555-123-4567  |                    | +1 555-867-5309  |
+------------------+                    +------------------+
        |                                      |
        +--------------------------------------+
              Same structure, zero PII

3 commands to anonymized data:

# 1. Install
go install github.com/felipevolpatto/mimicry/cmd/mimicry@latest

# 2. Initialize config
mimicry init

# 3. Run
mimicry run

For a complete working example with sample data, see the demo.

Consider a production database with sensitive user information:

Original data (users table):

After running Mimicry:

Key observations:

• Names are replaced with generated names from a predefined pool
• Email addresses are anonymized while preserving the format
• Phone numbers maintain their format but digits are transformed
• Salaries are statistically blurred (gaussian) while preserving distribution
• Primary keys (id) remain unchanged to preserve referential integrity

The same transformation is applied consistently across all tables. If "John Smith" appears in both users and audit_logs, he becomes "Morgan Davis" in both places.

Foreign keys and joins work perfectly in the shadow database. The same person, order, or entity is transformed identically across all tables.

Automatically identifies sensitive columns by name patterns (email, phone, first_name, address, etc.) and applies appropriate transformations. Works globally with international data formats.

• Built-in: Email, Name, Phone, Address, Date, IP, UUID, Gaussian blur
• Configurable: YAML-based overrides per table/column
• Extensible: Plugin system for domain-specific logic

Gaussian blur for numeric data maintains distribution patterns while masking individual values — perfect for analytics testing.

Processes data as a stream, never loading entire tables into memory. Handles multi-terabyte databases without OOM errors.

Only anonymize changes since the last run — ideal for CI/CD pipelines.

Pull a "slice" of production: "Give me 10% of users but include all their related orders."

git clone https://github.com/felipevolpatto/mimicry.git
cd mimicry
go build -o mimicry ./cmd/mimicry

go install github.com/felipevolpatto/mimicry/cmd/mimicry@latest

Create mimicry.yaml:

# Secret salt for deterministic hashing
# IMPORTANT: Keep this secret and consistent between runs
salt: "your-secret-salt-keep-this-safe"

# Source database
source:
  driver: postgres
  host: localhost
  port: 5432
  database: myapp_production
  username: readonly_user
  password: ${DB_PASSWORD}  # Environment variable

# Destination
destination:
  type: csv
  path: ./anonymized_output

# Skip certain tables
exclude_tables:
  - schema_migrations
  - ar_internal_metadata

# Custom transformer overrides
transformers:
  # Keep IDs unchanged
  users.id:
    skip: true
  
  # Use specific transformer
  users.salary:
    transformer: gaussian
    options:
      variance: 0.15  # 15% variance
  
  # Preserve email domain for internal testing
  users.email:
    transformer: email
    options:
      preserve_domain: true

# Processing options
options:
  batch_size: 1000
  workers: 4

# Initialize configuration
mimicry init

# Run anonymization
mimicry run -c mimicry.yaml

# Inspect database schema
mimicry inspect

# Validate configuration
mimicry validate

# Verbose output
mimicry run -v

+--------------+     +------------------+     +-------------+
|   Inspector  |---->|   Transformer    |---->|    Sink     |
|              |     |     Engine       |     |             |
| - Schema     |     |                  |     | - CSV       |
| - FK detect  |     | - Deterministic  |     | - JSON      |
| - Stream     |     | - Parallel       |     | - Database  |
+--------------+     +------------------+     +-------------+
       |                     |                       |
       +---------------------+-----------------------+
                    Hasher (SHA-256 + Salt)

See ARCHITECTURE.md for detailed technical documentation.

MIT License - see LICENSE for details.