Due to the complexity of cybersecurity datasets and their missing taxonomy mappings it
is time consuming to lead from a real life vulnerability to its mitigation technique. The
aim of this thesis was to create a program that does these steps more time efficiently and
creates a visual roadmap using these information. From the inserted identifier of one of
the major cybersecurity dataset entries (Common Vulnerabilities and Exposures, Common Weaknesses Enumeration, Common Attack Pattern Enumeration and Classification,
MITRE ATT&CK and MITRE D3FEND), free text or a product, detailed information
is fetched. Where datasets do not provide direct links, the system uses a Natural Language Processing model to suggest additional connections. These steps are repeated for
every related item, creating a complete roadmap that shows how a vulnerability can lead
through different stages of exploitation and ultimately to applicable defense techniques.
The final output is displayed through a web-based interactive graph that allows users
to explore the data, move nodes, view brief descriptions, and open detailed JSON files.
Testing with various input types showed that the program handles all valid identifiers
correctly and produces accurate results. Future improvements include speeding up the
NLP component, improving its accuracy and adding support for searching by keywords.
To support reproducibility and further development, the complete source code and application are publicly accessible in the designated GitHub repository via the following link:
https://github.com/RumanRobert/VEACOR.git.