A privacy-focused document redaction tool that automatically detects and redacts sensitive information from PDF and DOCX documents.

🔒 100% Local Processing • 🚀 No Server Required • 🎯 Smart Detection • 🛡️ Privacy First

• Automatic Detection: Uses regex patterns and ML-based detection to find sensitive information
• Multiple Detection Methods: Combines pattern matching with machine learning for high accuracy
• Customizable Patterns: Add your own predefined words and custom detection rules
• Language Support: Detects content in multiple languages (Latin, Chinese, Arabic, etc.)

• PDF Files: Full support for text-based PDFs
• DOCX Files: Microsoft Word document support
• Metadata Extraction: View and remove document metadata and hidden content
• Image Extraction: Extract and download embedded images from documents

• Interactive Review: Review and confirm detected entities before redaction
• Manual Selection: Click and drag to manually select sensitive areas
• Bulk Operations: Confirm or reject multiple entities at once
• Real-time Preview: See exactly what will be redacted

• 100% Local Processing: All processing happens in your browser
• No Server Required: No data is sent to external servers
• No Tracking: No analytics or telemetry
• Open Source: Fully auditable code

Safe Redact automatically detects the following types of sensitive information in both PDF and DOCX documents:

• Social Security Numbers (SSN): US Social Security Numbers in various formats
• Email Addresses: RFC 5322 compliant email addresses
• Phone Numbers:
  • US phone numbers (various formats)
  • International phone numbers
  • China mobile numbers (11-digit format)
  • China landline numbers

• Credit Card Numbers:
  • Generic credit cards (13-19 digits with Luhn validation)
  • Visa cards (13 or 16 digits)
  • Mastercard (16 digits)
  • American Express (15 digits)
  • Discover cards
  • China UnionPay cards

• MM/DD/YYYY format
• YYYY-MM-DD format
• Month DD, YYYY format
• Chinese date format (YYYY年MM月DD日)
• DD/MM/YYYY format (International)

• Chinese National ID: 18-digit ID with validation
• Passports:
  • Chinese Passport (current and legacy formats)
  • US Passport
• Network Information:
  • IPv4 addresses
  • IPv6 addresses
  • URLs (HTTP/HTTPS)
• Cryptocurrency: Bitcoin addresses
• Custom Predefined Words: User-defined sensitive terms

When the "Sanitize Document" option is enabled, Safe Redact removes metadata and hidden content to prevent information leakage.

The following elements are removed from PDF documents:

Metadata:

• Title, Author, Subject, Keywords
• Creator, Producer
• Creation Date, Modification Date
• All other metadata fields

Hidden Content:

• Comments and annotations (all types)
• Markup annotations (highlights, underlines, strikeouts, etc.)
• Stamps and file attachments
• Multimedia content (sound, video)
• Form fields (optional)
• Embedded files
• JavaScript actions
• Optional Content Groups (PDF layers)

The following elements are removed from DOCX documents:

Metadata:

• Core Properties: Title, Author, Subject, Keywords, Creator, Last Modified By, Dates, Category, Content Status
• App Properties: Application name, Version, Company, Manager, Template
• Custom Properties: All custom metadata

Hidden Content:

• Comments and comment references
• Track Changes/Revisions (insertions, deletions, moves, formatting changes)
• Bookmarks
• Custom XML data
• Document Settings: Revision identifiers (RSIDs), proof errors, document protection
• VBA/Macros and macro data
• (Optional) Headers and footers
• (Optional) Embedded objects and files

• Node.js 18 or higher
• npm or yarn

1. Clone the repository:

git clone https://github.com/zhendong/safe-redact.git
cd safe-redact

2. Install dependencies:

npm install

Start the development server:

npm run dev

The application will be available at http://localhost:5173

Build for production:

npm run build

The built files will be in the dist/ directory.

Run the test suite:

npm test

• React 19 + TypeScript: UI framework with type safety
• Vite: Fast build tool and dev server
• MuPDF: PDF parsing, rendering, and manipulation
• PizZip + Mammoth: DOCX processing
• Transformers.js: ML-based entity detection (optional)
• Tailwind CSS: Utility-first styling

• All document processing happens 100% locally in your browser
• No data is sent to external servers
• Documents never leave your device
• Open source and auditable

Contributions are welcome!

• Report bugs and issues
• Suggest new features
• Improve documentation
• Submit pull requests
• Add test cases

MIT License - see LICENSE file for details