Prompt Injection Defense System

Design a defense system against prompt injection attacks.

## Application Context
{{application_context}}

## User Input Sources
{{input_sources}}

## Risk Tolerance
{{risk_tolerance}}

Implement layered defenses:

**Layer 1: Input Sanitization**
- Pattern matching for known attacks
- Unicode normalization
- Special character handling

**Layer 2: Prompt Structure**
- Input/instruction separation
- Privileged context isolation
- Output format constraints

**Layer 3: Output Validation**
- Response anomaly detection
- Sensitive data leakage prevention
- Action verification

**Layer 4: Monitoring**
- Attack attempt logging
- Behavioral anomaly detection
- Incident response triggers

Provide:
- Implementation code for each layer
- Testing methodology
- False positive management