A web application for comparing LLM and knowledge graph performance on cybersecurity queries

The evolution of cybersecurity has led to a spike in digital threats, both in frequency and complexity, necessitating advanced, intelligent solutions to protect sensitive information. Traditional defense mechanisms are increasingly inadequate, pushing cybersecurity professionals to seek innovative approaches for threat detection, response, and data analysis. This thesis investigates the integration of Large Language Models (LLMs) and Knowledge Graphs into cybersecurity workflows to address these challenges. Specifically, it explores the development of a web application that enables real-time, interactive use of state-of-the-art LLMs, such as OpenAI’s GPT-4 and similar models, for improved threat response and workflow efficiency. Built with a React frontend and FastAPI backend, the application allows for seamless interactions with multiple LLMs, offering tools to evaluate model responses, track performance, and handle cybersecurity-specific queries. The inclusion of Knowledge Graphs further improves the structured retrieval of information, providing cybersecurity professionals with a platform for managing complex cybersecurity efforts. Additionally, an automated performance evaluation system ensures response accuracy, crucial for sensitive cybersecurity tasks. This research demonstrates the potential of LLMs to benefit cybersecurity capabilities, showing their role in advancing threat detection, response generation, and comprehensive data analysis.