KoreShield is a security platform designed to protect enterprise applications that use Large Language Models (LLMs) from prompt injection attacks. It sits transparently between your application and LLM API providers (OpenAI, Anthropic, Google Gemini, etc.), sanitizing inputs, detecting threats, and enforcing security policies before requests reach the model.

This project is unified under the KoreShield brand, with all resources, SDKs, and documentation available at koreshield.ai.

The goal is for KoreShield to be the "security layer" for any LLM-powered enterprise system—easy to integrate, highly configurable, and always up-to-date with the latest defenses and best practices.

• LLM-Firewall: A Lightweight Middleware Architecture for Real-Time Detection of Indirect Prompt Injection in Enterprise RAG System

• Preprint: LLM Firewall – A Novel Taxonomy of Indirect Prompt Injection Attacks in Enterprise RAG Systems

www.koreshield.ai