Skip to main content

​
Features overview

KoreShield provides comprehensive security protection for LLM applications through multiple layers of defense. Explore our key features below:

RAG Defense Engine

Detect and block indirect prompt injection in RAG systems

Attack Detection

Multi-layered detection of prompt injection and security threats

Security Policies

Configurable policies for threat response and enforcement

Input Sanitization

Clean and normalize prompts to remove malicious content

​
Core Security Capabilities

KoreShield protects your LLM applications through four core security components:

​
Input Sanitization

Cleans and normalizes prompts before they reach the LLM:
  • Removes potentially malicious content patterns
  • Normalizes encoding and special characters
  • Strips dangerous formatting attempts
  • Validates input structure

​
Attack Detection

Analyzes prompts and responses for signs of attacks:
  • Keyword-Based Detection - Direct injection phrases and patterns
  • Pattern Analysis - Code injection, role manipulation, encoding tricks
  • Custom Rules - Flexible rule engine for your specific needs
  • ML-Inspired Heuristics - Anomaly detection and scoring

​
Policy Enforcement

Applies configurable security rules:
  • Sensitivity levels (low, medium, high)
  • Response actions (allow, warn, block)
  • Custom policies per environment
  • Allowlists and blocklists

​
Audit Logging

Records all security events and decisions:
  • Comprehensive security event tracking
  • Compliance-ready audit trails
  • Real-time monitoring integration
  • Detailed threat taxonomy

​
Multi-Provider Support

KoreShield works seamlessly with all major LLM providers:

OpenAI

GPT-3.5, GPT-4, and all models

Anthropic

Claude 3.5 Sonnet and family

DeepSeek

High-performance models

Google Gemini

Gemini Pro and Ultra

Azure OpenAI

Enterprise OpenAI deployment

Custom Models

Any OpenAI-compatible API

​
Key Benefits

Sub-millisecond latency with comprehensive security scanning. KoreShield adds minimal overhead to your LLM requests while providing complete protection.
Drop-in replacement for existing LLM API calls. Simply point your application to the KoreShield proxy and you’re protected.
  • SOC 2 Type II compliant
  • GDPR and HIPAA ready
  • Comprehensive audit trails
  • Multi-tenant support
Transparent security with community-driven improvements. Audit the code, contribute features, and customize for your needs.

​
Performance Metrics

  • Latency: P50 < 50ms, P99 < 300ms
  • Throughput: 1000+ requests/second per instance
  • Detection Accuracy: 99.5% true positive rate
  • False Positive Rate: < 0.1% with default settings

​
Next Steps

Try RAG Defense

Protect your RAG pipelines from indirect injection

Configure Policies

Set up security policies for your environment

View Integrations

Explore framework and platform integrations

API Reference

Dive into the complete API documentation

Build docs developers (and LLMs) love

Get started for freeTalk to us