LLMGoat

KO | EN

gitlite — search

#bash #nlp #macos #javascript #claude-code #sql #claude #ui #anthropic #golang

(★ 9)

No description available.

File Explorer

ZIP Download

- - __init__.py
- Dockerfile
- llm_engine.py
- main.py
- requirements.txt
- app.js
- Dockerfile
- index.html
.env.example
.gitignore
docker-compose.yml
README.md

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

# Project Badges

Docker
FastAPI
Python
CyberSecurity

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

**LLMGoat** is an intentionally vulnerable environment designed for security researchers, pentesters, and developers to practice, test, and understand security vulnerabilities present in **Large Language Model (LLM)** applications based on the **OWASP Top 10 for LLMs** framework.

---

## 🚀 Covered Vulnerabilities (OWASP LLM Top 10)

| ID | Vulnerability | Description |
|---|---|---|
| **LLM01** | **Prompt Injection** | Craft adversarial inputs to bypass system guardrails and alter execution behavior. |
| **LLM02** | **Sensitive Information Disclosure** | Extract confidential system prompts, secret API keys, or hidden PII data. |
| **LLM03** | **Supply Chain Vulnerabilities** | Simulate risks from third-party plugins, compromised packages, or untrusted models. |
| **LLM04** | **Data & Model Poisoning** | Manipulate training data, fine-tuning datasets, or RAG ingestion sources. |
| **LLM05** | **Improper Output Handling** | Exploit unformatted or unvalidated LLM output leading to XSS or RCE. |
| **LLM06** | **Excessive Agency** | Abuse autonomous agent permissions, tooling rights, and administrative capabilities. |
| **LLM07** | **System Prompt Leakage** | Trick the LLM into disclosing hidden operational instructions and context rules. |
| **LLM08** | **Vector & Embedding Weaknesses** | Exploit flaws in Retrieval-Augmented Generation (RAG) vector retrieval logic. |
| **LLM09** | **Misconfiguration & Overreliance** | Leverage hallucinated endpoints, improper default settings, and loose access bounds. |
| **LLM10** | **Model Theft** | Perform side-channel attacks or prompt reverse-engineering to extract model logic. |

---

## 🛠️ Architecture Overview

- **Frontend**: Clean interactive UI built to test prompts across different LLM challenge modules.
- **Backend**: Python FastAPI microservice handling LLM engine simulations, request evaluations, and flag verifications.
- **Containerization**: Fully modular setup packaged via Docker & Docker Compose for hassle-free deployment.

---

## ⚡ Quick Start & Installation

### Prerequisites
- [Docker](https://docs.docker.com/get-docker/)
- [Docker Compose](https://docs.docker.com/compose/install/)
- [Git](https://git-scm.com/)

### 1️⃣ Clone the Repository
```bash
git clone git@github.com:Nodevik/LLMGoat.git
cd LLMGoat
```

### 2️⃣ Environment Setup
Copy the template environment file to `.env`:
```bash
cp .env.example .env
```

### 3️⃣ Launch with Docker Compose
```bash
docker compose up --build -d
```

Access the application in your browser:
- 🌐 **Frontend Interface**: `http://localhost:80`
- ⚙️ **Backend API Specs**: `http://localhost:8000/docs`

---

## 🛡️ Disclaimer

This application contains intentionally vulnerable code designed strictly for **educational, security research, and authorized penetration testing purposes**. Do not deploy this project onto public or production infrastructure.

---

## 🤝 Author & Connect

Created and maintained by **Vikas Vaishnav**.

Connect with me for cybersecurity research, LLM security discussions, and project updates:

- 💼 **LinkedIn**: [Vikas Vaishnav](https://www.linkedin.com/in/vikas-vaishnav-0719262b8/)
- 📷 **Instagram**: [@_cyberwizard_](https://www.instagram.com/_cyberwizard_/)
- 🐙 **GitHub**: [@Nodevik](https://github.com/Nodevik)

LLMGoat

Owner Repos

File Explorer

# Use via CDN

# Project Badges

Was this content helpful?

Recent Feedback

LLMGoat

Owner Repos

File Explorer

# Use via CDN

# Project Badges

Was this content helpful?

Recent Feedback

Download README