VOSviewer Program for Clustering and Visualizing Cyber Threat Information

Abstract

This study considers a method for classifying vulnerabilities and cyber threats using data from information and news resources. The authors propose to apply text analysis and machine learning methods to automate the detection and classification of threats in information systems. Various approaches are explored, including latent semantic analysis, topic modeling, n-gram analysis, and vector representation of text, which allows identifying semantic relationships and thematic structures. These methods help to adapt approaches to the requirements for interpretability and accuracy of cyber threat data. The paper considers an approach to analyzing the co-occurrence of words and phrases in texts to identify thematic clusters of cyber threats. The main analysis tool is the calculation of the association measure between terms, which allows for a quantitative assessment of their relationships and contributes to the construction of more effective classification models using machine learning methods. To visualize the results, it is proposed to use strategic diagrams based on two key indicators: centrality and cluster density. Clusters are classified into four categories depending on the values of these indicators, which allows for a more accurate characterization of their essence and relationships with other threats. To analyze the dynamics of changes in clusters, the method of directed graphs is used, which allows tracking the transformation of cyber threat components from one time period to another. Based on the presented approach, it is possible to create more informative cyber threat monitoring systems, which ultimately contributes to increasing the level of protection of information systems from potential attacks. Thus, the study opens up new horizons for automating the analysis and classification of threats, providing more effective solutions in the field of cybersecurity.