Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 

Repository files navigation

🧠 AI Hybrid Red Teaming: When Traditional Web Vulnerabilities Meet LLM-Driven Systems

A hands-on security lab demonstrating how traditional web vulnerabilities can be combined with modern AI systems in realistic application environments.


⚠️ Disclaimer

This is a personal project created for educational purposes.

  • All scenarios, data, and interactions are fully synthetic and created from scratch
  • This project is intentionally vulnerable and must not be used in production environments

🧩 Challenge

You are interacting with an internal-style assistant used for engineering support and troubleshooting.

Explore the application and understand how it behaves.


🎯 Objective

Identify and retrieve the credential.


💡 Hint

  • Observe how the application communicates with the backend
  • Pay attention to how conversations are handled
  • Think beyond the UI

🧠 What This Lab Covers

This lab focuses on the intersection of:

  • Traditional web security issues
  • AI-assisted application behavior
  • Realistic user interaction patterns

🛠️ Tech Stack

  • Backend: FastAPI
  • Database: SQLite
  • Frontend: HTML / CSS / JavaScript
  • LLM: Ollama (local model)

⚙️ Prerequisites


🚀 Setup & Run

git clone https://github.com/hoseinvita1337/VitAILab.git
cd backend

python3 -m venv .venv
source .venv/bin/activate

pip install -r requirements.txt

uvicorn app.main:app --reload

🤖 Model Setup

ollama serve
# in another tab
ollama pull phi3:mini

Run as a container

#OLLAMA_HOST=0.0.0.0:11434 ollama serve #Make sure ollama is accessible
#systemctl start docker #Make sure docker is running
docker build -t vitailab .
docker run -p 8000:8000 --add-host=host-gateway:host-gateway vitailab
#docker run -p 8000:8000 -e OLLAMA_URL=http://host.docker.internal:11434 vitailab #MacOS & Win

🌐 Access

http://localhost:8000

🔐 Test Users

Username Password

user2 password123


🧪 Notes

The application is intentionally designed with insecure behavior The goal is to explore and understand the interaction between components


📁 Project Structure

├── backend
│   ├── app
│   │   ├── auth.py
│   │   ├── chat.py
│   │   ├── config.py
│   │   ├── db.py
│   │   ├── llm.py
│   │   ├── main.py
│   │   ├── models.py
│   │   └── static
│   │       └── index.html
│   ├── lab.db
│   └── requirements.txt
└── README.md

👤 Author

Hosein Vita

About

AI hybrid red teaming lab demonstrating how traditional web vulnerabilities (IDOR) can be chained with LLM behavior to expose sensitive data in modern applications.

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages