LLM Proxy PII Masking

Automatically detect and redact personally identifiable information before sending requests to AI models. Protect sensitive data, ensure privacy compliance, and prevent data leakage.

🔒 Auto Detection 🎭 Smart Masking 📋 GDPR Ready ⚡ Real-time
🎭

PII Masking Demo

Protected
Original Request
Hi, my name is John Smith and my email is john.smith@company.com. My SSN is 123-45-6789 and I need help with my account ending in 4521.
Masked Request (Sent to AI)
Hi, my name is [NAME] and my email is [EMAIL]. My SSN is [SSN] and I need help with my account ending in [ACCOUNT].
4
PII Detected
100%
Detection Rate
<5ms
Processing Time
0
Data Leaked

PII Masking Features

Comprehensive privacy protection for your LLM API communications

🔍

Automatic Detection

AI-powered detection identifies PII in real-time using pattern matching, NER models, and custom rules for accurate identification.

🎭

Smart Masking

Replace sensitive data with contextually appropriate placeholders that preserve meaning while protecting privacy.

🔄

Reversible Masking

Optionally store mappings to restore original values in responses when needed for specific use cases.

⚙️

Custom Rules

Define custom PII patterns and masking rules specific to your organization's data types and requirements.

📊

Audit Logging

Complete audit trail of all PII detection and masking events for compliance and security analysis.

Zero Latency

Optimized processing pipeline adds minimal overhead to request latency, typically under 5 milliseconds.

PII Types Detected

Automatically identify and protect various categories of sensitive information

👤

Names

First names, last names, full names of individuals

📧

Email Addresses

Personal and business email addresses

📞

Phone Numbers

Mobile, landline, and international numbers

🏠

Addresses

Street addresses, cities, postal codes

🆔

SSN / Tax IDs

Social security numbers, tax identification

💳

Payment Data

Credit card numbers, bank accounts

🏥

Medical Info

Health IDs, medical record numbers

🌐

IP Addresses

IPv4 and IPv6 addresses

How It Works

Four-step process for comprehensive PII protection

1

Intercept Request

Proxy intercepts outgoing request to LLM API before it leaves your infrastructure.

2

Detect PII

Scan request content using pattern matching and ML models to identify all PII.

3

Apply Masking

Replace detected PII with appropriate placeholders while preserving context.

4

Forward Safely

Send masked request to AI model with sensitive data protected.

Compliance Standards

GDPR

GDPR Compliant

European data protection regulation compliance

HIPAA

HIPAA Ready

Healthcare data privacy protection

PCI

PCI DSS

Payment card data security standards

CCPA

CCPA Compliant

California Consumer Privacy Act

Configuration Example

pii_masking_config.yaml
# PII Masking Configuration
pii_masking:
  enabled: true
  
  detection:
    methods:
      - "regex"           # Pattern matching
      - "ner"            # Named entity recognition
      - "custom_rules"   # Organization-specific rules
    
    entity_types:
      - "PERSON"          # Names
      - "EMAIL"          # Email addresses
      - "PHONE"          # Phone numbers
      - "SSN"            # Social security numbers
      - "CREDIT_CARD"    # Payment card numbers
      - "ADDRESS"        # Physical addresses
  
  masking:
    strategy: "placeholder"
    placeholders:
      PERSON: "[NAME]"
      EMAIL: "[EMAIL]"
      PHONE: "[PHONE]"
      SSN: "[SSN]"
    
    reversible: true
    storage: "encrypted_vault"

Related Resources

LLM Proxy IP Whitelist

Combine IP restrictions with PII masking for comprehensive data protection.

LLM Proxy Usage Analytics

Track PII detection statistics and masking events in your analytics dashboard.

LLM Proxy Multi-Provider Routing

Apply consistent PII masking across all LLM providers seamlessly.

LLM Proxy Cost Tracking Dashboard

Monitor costs while ensuring PII never leaves your infrastructure.

Protect Your Sensitive Data

Implement automatic PII masking and ensure privacy compliance for all your LLM API communications.