// LLM · Capstone Project

ChatCSEC

A cybersecurity chatbot leveraging OpenAI's GPT model, integrating private or internal document embeddings for accurate, contextual answers to security questions. Collaboratively developed by Domenic Lo Iacono, Kyri Lea, Brian McNulty, and Rich Kleinhenz.

View on GitHub ← Back to portfolio

Project Overview

The ChatCSEC application includes modular components such as a user interface, AI models, embedding systems, a vector database, and a web scraper. Together, these components create a robust platform for cybersecurity knowledge sharing and query resolution.

Application Components

User Interface: The primary interface for interacting with ChatCSEC.
Model: Utilizes OpenAI's GPT-4-0125-preview model to formulate responses. Requires an OpenAI API key.
Embedding Model: Processes documents using OpenAI's text-embedding-3-small model to extend the chatbot's knowledge base. Requires an OpenAI API key.
Vector Database: QDrant stores embeddings for efficient retrieval and training persistence.
Web Scraper: Crawls websites to gather content and embed new data, ensuring the chatbot remains up-to-date with recent information.

Key Features

Cybersecurity Expertise: Provides reliable and accurate responses tailored to cybersecurity topics.
Customizable Knowledge Base: Integrates private documents into the model's training data via embeddings.
Automated Updates: Includes a web scraper for dynamic updates, ensuring information remains current.
Modular Design: Flexible architecture for substituting components such as models, databases, or scrapers.

Installation

To set up ChatCSEC, developers should:

Set up an OpenAI account and obtain API keys for the specified models.
Install QDrant as the vector database.
Run the application locally and configure component connections as needed.

Usage

ChatCSEC allows users to interact with a cybersecurity chatbot and train it using internal or external documents. Documentation generation is supported using Sphinx. To build documentation:

Run sphinx-apidoc -o docs ChatCSEC to create or update the rst files.
Use the Makefile in the docs directory to generate HTML, Latex, or other formats. The generated files will appear in /docs/_build/.

Future Work

Expand compatibility with additional vector databases and embedding models.
Improve the web scraper to handle more complex websites and file formats.
Introduce a more user-friendly interface for non-technical users.
Enable advanced query analytics to identify trends in cybersecurity queries.