@himorishige/noren-plugin-jp

Japanese PII detection plugin for Noren

Specialized detectors and maskers for Japanese Personally Identifiable Information (PII) including phone numbers, postal codes, and My Number (マイナンバー).

✨ Features

📞 Phone Numbers

• Mobile: 090, 080, 070, 060 prefixes
• Landline: Area codes with proper validation
• International: +81 format support
• Context-aware: Uses hints like "電話", "TEL"

📮 Postal Codes

• Standard format: 123-4567
• Numeric format: 1234567
• Context detection: "〒", "住所" hints
• Smart masking: 〒123-4567 → 〒•••-••••
• ⚡ v0.6.0: Improved phone number conflict resolution

🆔 My Number (マイナンバー)

• 12-digit validation: Proper checksum verification
• Context required: Only detects with "マイナンバー", "個人番号" hints
• High security: [REDACTED:MYNUMBER] output

🚀 Installation

npm install @himorishige/noren-plugin-jp @himorishige/noren-core

Basic Usage

import { Registry, redactText } from '@himorishige/noren-core';
import * as jpPlugin from '@himorishige/noren-plugin-jp';

// Initialize the Registry
const registry = new Registry({
  defaultAction: 'mask',
  enableConfidenceScoring: true, // v0.4.0: Improved accuracy
  environment: 'production',
  // Set relevant keywords as hints to improve detection accuracy
  contextHints: ['電話', '住所', '〒', 'マイナンバー'],
});

// Register the detectors and maskers from the Japan plugin
registry.use(jpPlugin.detectors, jpPlugin.maskers);

const inputText = 'My phone number is 090-1234-5678, and my address is 〒150-0001.';

// Execute the redaction process
const redactedText = await redactText(registry, inputText);

console.log(redactedText);
// Output: My phone number is •••-••••-••••, and my address is 〒•••-••••.

Detected Types

| PII Type | Description | Masking Example (mask) | v0.5.0 | | :------------- | :------------------- | :----------------------- | :------ | | phone_jp | Japanese phone number| •••-••••-•••• | ✓ Renamed | | postal_jp | Japanese postal code | 〒•••-•••• | ✓ Renamed | | mynumber_jp | My Number | [REDACTED:MYNUMBER] | ✓ Renamed |

Full-Width Character Support

This plugin works seamlessly with full-width (zenkaku) characters through Noren's built-in Unicode NFKC normalization:

// Both half-width and full-width characters are detected equally
const halfWidth = '電話: 090-1234-5678'
const fullWidth = '電話: ０９０-１２３４-５６７８'

// Both produce identical masking results
const result1 = await redactText(registry, halfWidth)  // → 電話: •••-••••-••••
const result2 = await redactText(registry, fullWidth)  // → 電話: •••-••••-••••

🆕 What's New in v0.6.0

⚡ Enhanced Postal Code Detection

• Phone Number Conflict Resolution: Postal codes are no longer misdetected when they match phone number patterns
• Improved Accuracy: TEL: 03-1234-5678 is correctly identified as a phone number, not a postal code
• Lightweight Implementation: Simplified algorithm for better performance while maintaining accuracy

🔧 Technical Improvements

• Removed heavy address dictionary dependencies
• Streamlined detection logic focusing on core functionality
• Maintained Noren's "Edge-native, Lightweight" design principles

📊 Performance Characteristics

• Fast Processing: Optimized for edge environments
• Low Memory Usage: Minimal memory footprint
• Stream-First: Designed for efficient streaming operations

Examples of Improved Detection

// v0.6.0 correctly distinguishes between phone and postal codes
const testCases = [
  'TEL: 03-1234-5678',           // → Detected as phone_jp
  '〒100-0001 東京都千代田区',     // → Detected as postal_jp (high confidence)
  '住所: 123-4567',              // → Detected as postal_jp (medium confidence)
  'Phone: 090-1111-2222',        // → Detected as phone_jp
]