DataFog: PII Detection & Anonymization

Fast processing • Production-ready • Simple configuration

---

Overview

DataFog provides efficient PII detection using a pattern-first approach that processes text significantly faster than traditional NLP methods while maintaining high accuracy.

# Basic usage example
from datafog import DataFog
results = DataFog().scan_text("John's email is [email protected] and SSN is 123-45-6789")

Performance Comparison

Engine	10KB Text Processing	Relative Speed	Accuracy
DataFog (Regex)	~2.4ms	190x faster	High (structured)
DataFog (GLiNER)	~15ms	32x faster	Very High
DataFog (Smart)	~3-15ms	60x faster	Highest
spaCy	~459ms	baseline	Good

Performance measured on 13.3KB business document. GLiNER provides excellent accuracy for named entities while maintaining speed advantage.

Supported PII Types

Type	Examples	Use Cases
Email	[email protected]	Contact scrubbing
Phone	(555) 123-4567	Call log anonymization
SSN	123-45-6789	HR data protection
Credit Cards	4111-1111-1111-1111	Payment processing
IP Addresses	192.168.1.1	Network log cleaning
Dates	01/01/1990	Birthdate removal
ZIP Codes	12345-6789	Location anonymization

---

Quick Start

Installation

# Lightweight core (fast regex-based PII detection)
pip install datafog

# With advanced ML models for better accuracy
pip install datafog[nlp]                # spaCy for advanced NLP
pip install datafog[nlp-advanced]       # GLiNER for modern NER
pip install datafog[ocr]                # Image processing with OCR
pip install datafog[all]                # Everything included

Basic Usage

Detect PII in text:

from datafog import DataFog

# Simple detection (uses fast regex engine)
detector = DataFog()
text = "Contact John Doe at [email protected] or (555) 123-4567"
results = detector.scan_text(text)
print(results)
# Finds: emails, phone numbers, and more

# Modern NER with GLiNER (requires: pip install datafog[nlp-advanced])
from datafog.services import TextService
gliner_service = TextService(engine="gliner")
result = gliner_service.annotate_text_sync("Dr. John Smith works at General Hospital")
# Detects: PERSON, ORGANIZATION with high accuracy

# Best of both worlds: Smart cascading (recommended for production)
smart_service = TextService(engine="smart")
result = smart_service.annotate_text_sync("Contact [email protected] or call (555) 123-4567")
# Uses regex for structured PII (fast), GLiNER for entities (accurate)

Anonymize on the fly:

# Redact sensitive data
redacted = DataFog(operations=["scan", "redact"]).process_text(
    "My SSN is 123-45-6789 and email is [email protected]"
)
print(redacted)
# Output: "My SSN is [REDACTED] and email is [REDACTED]"

# Replace with fake data
replaced = DataFog(operations=["scan", "replace"]).process_text(
    "Call me at (555) 123-4567"
)
print(replaced)
# Output: "Call me at [PHONE_A1B2C3]"

Process images with OCR:

import asyncio
from datafog import DataFog

async def scan_document():
    ocr_scanner = DataFog(operations=["extract", "scan"])
    results = await ocr_scanner.run_ocr_pipeline([
        "https://example.com/document.png"
    ])
    return results

# Extract text and find PII in images
results = asyncio.run(scan_document())

---

Advanced Features

Engine Selection

Choose the appropriate engine for your needs:

from datafog.services import TextService

# Regex: Fast, pattern-based (recommended for speed)
regex_service = TextService(engine="regex")

# spaCy: Traditional NLP with broad entity recognition
spacy_service = TextService(engine="spacy")

# GLiNER: Modern ML model optimized for NER (requires nlp-advanced extra)
gliner_service = TextService(engine="gliner")

# Smart: Cascading approach - regex → GLiNER → spaCy (best accuracy/speed balance)
smart_service = TextService(engine="smart")

# Auto: Regex → spaCy fallback (legacy)
auto_service = TextService(engine="auto")

Performance & Accuracy Guide:

Engine	Speed	Accuracy	Use Case	Install Requirements
regex	🚀 Fastest	Good	Structured PII (emails, phones)	Core only
gliner	⚡ Fast	Better	Modern NER, custom entities	pip install datafog[nlp-advanced]
spacy	🐌 Slower	Good	Traditional NLP entities	pip install datafog[nlp]
smart	⚡ Balanced	Best	Combines all approaches	pip install datafog[nlp-advanced]

Model Management:

# Download specific GLiNER models
import subprocess

# PII-specialized model (recommended)
subprocess.run(["datafog", "download-model", "urchade/gliner_multi_pii-v1", "--engine", "gliner"])

# General-purpose model
subprocess.run(["datafog", "download-model", "urchade/gliner_base", "--engine", "gliner"])

# List available models
subprocess.run(["datafog", "list-models", "--engine", "gliner"])

Anonymization Options

from datafog import DataFog
from datafog.models.anonymizer import AnonymizerType, HashType

# Hash with different algorithms
hasher = DataFog(
    operations=["scan", "hash"],
    hash_type=HashType.SHA256  # or MD5, SHA3_256
)

# Target specific entity types only
selective = DataFog(
    operations=["scan", "redact"],
    entities=["EMAIL", "PHONE"]  # Only process these types
)

Batch Processing

documents = [
    "Document 1 with PII...",
    "Document 2 with more data...",
    "Document 3..."
]

# Process multiple documents efficiently
results = DataFog().batch_process(documents)

---

Performance Benchmarks

Performance comparison with alternatives:

Speed Comparison (10KB text)

DataFog Pattern:  4ms   ████████████████████████████████ 123x faster
spaCy:         480ms   ██ baseline

Engine Selection Guide

Scenario	Recommended Engine	Why
High-volume processing	pattern	Maximum speed, consistent performance
Unknown entity types	spacy	Broader entity recognition
General purpose	auto	Smart fallback, best of both worlds
Real-time applications	pattern	Sub-millisecond processing

---

CLI Usage

DataFog includes a command-line interface:

# Scan text for PII
datafog scan-text "John's email is [email protected]"

# Process images
datafog scan-image document.png --operations extract,scan

# Anonymize data
datafog redact-text "My phone is (555) 123-4567"
datafog replace-text "SSN: 123-45-6789"
datafog hash-text "Email: [email protected]" --hash-type sha256

# Utility commands
datafog health
datafog list-entities
datafog show-config

---

Features

Security & Compliance

• Detection of regulated data types for GDPR/CCPA compliance
• Audit trails for tracking detection and anonymization
• Configurable detection thresholds

Scalability

• Batch processing for handling multiple documents
• Memory-efficient processing for large files
• Async support for non-blocking operations

Integration Example

# FastAPI middleware example
from fastapi import FastAPI
from datafog import DataFog

app = FastAPI()
detector = DataFog()

@app.middleware("http")
async def redact_pii_middleware(request, call_next):
    # Automatically scan/redact request data
    pass

---

Common Use Cases

Enterprise

• Log sanitization
• Data migration with PII handling
• Compliance reporting and audits

Data Science

• Dataset preparation and anonymization
• Privacy-preserving analytics
• Research compliance

Development

• Test data generation
• Code review for PII detection
• API security validation

---

Installation & Setup

Basic Installation

pip install datafog

Development Setup

git clone https://github.com/datafog/datafog-python
cd datafog-python
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements-dev.txt
just setup

Docker Usage

FROM python:3.10-slim
RUN pip install datafog
COPY . .
CMD ["python", "your_script.py"]

---

Contributing

Contributions are welcome in the form of:

• Bug reports
• Feature requests
• Documentation improvements
• New pattern patterns for PII detection
• Performance improvements

Quick Contribution Guide

# Setup development environment
git clone https://github.com/datafog/datafog-python
cd datafog-python
just setup

# Run tests
just test

# Format code
just format

# Submit PR
git checkout -b feature/your-improvement
# Make your changes
git commit -m "Add your improvement"
git push origin feature/your-improvement

See CONTRIBUTING.md for detailed guidelines.

---

Benchmarking & Performance

Run Benchmarks Locally

# Install benchmark dependencies
pip install pytest-benchmark

# Run performance tests
pytest tests/benchmark_text_service.py -v

# Compare with baseline
python scripts/run_benchmark_locally.sh

Continuous Performance Monitoring

Our CI pipeline:

• Runs benchmarks on every PR
• Compares against baseline performance
• Fails builds if performance degrades >10%
• Tracks performance trends over time

---

Documentation & Support

Resource	Link
Documentation	docs.datafog.ai
Community Discord	Join here
Bug Reports	GitHub Issues
Feature Requests	GitHub Discussions
Support	[email protected]

---

License & Acknowledgments

DataFog is released under the MIT License.

Built with:

• Pattern optimization for efficient processing
• spaCy integration for NLP capabilities
• Tesseract & Donut for OCR capabilities
• Pydantic for data validation

---

GitHub • Documentation • Discord