Možnosti chatbotů v kybernetické bezpečnosti

Abstract

Cybersecurity is facing a growing number and increasing sophistication of threats, along with rising demands for the protection of systems and organizational data. General-purpose large language models have limited applicability in this domain due to outdated knowledge, lack of access to internal data, and potential security risks.

This thesis focuses on the design and implementation of a specialized on-premise cybersecurity assistant. The solution combines a local language model with external knowledge base through a Retrieval-Augmented Generation approach over a vector database. The knowledge is based on authoritative cybersecurity sources.

The proposed system was experimentally evaluated and the results show that integrating internal knowledge significantly improves the quality of generated responses. The thesis demonstrates that a private, specialized assistant represents a practical and viable solution.