APPLICATION OF GENERATIVE ARTIFICIAL INTELLIGENCE FOR DETERMINE THE CENTROIDS AND ESSENCE OF CLUSTERS IN THE NETWORK OF CONCEPTS IN THE FIELD OF CYBERSECURITY

DOI: 10.31673/2412-4338.2025.012993

Abstract

Abstract. The complexity of the information environment requires new approaches to analyzing and classifying cyber threats. One of the promising methods is the study of networks of concepts in cybersecurity, identifying key concepts and relationships between key objects and phenomena in this area. Given the strong development of artificial intelligence technologies, it is quite reasonable to use modern large linguistic models that can be considered as "virtual experts" used under the guidance of a human analyst. The article presents a methodology for determining centroids and the essence of clusters in a network of cybersecurity objects based on generative artificial intelligence. The methodology is based on building a semantic network of concepts and clustering the network using the modularity algorithm. The procedure for determining the modularity classes is carried out according to an algorithm that includes the steps of initialization, evaluation, cluster merging, iteration, and finalization. The task of determining the centroids of concepts and the essence of clusters is delegated to a swarm of virtual experts - generative language models (GPT-4, Llama-3, o1), which allows automating and significantly speeding up this process. The methodology is illustrated by analyzing weekly cybersecurity digests. The results show that the use of a swarm of virtual experts can effectively automate the process of analyzing and classifying concepts in the network, which can serve as a basis for further improvement of cybersecurity classifiers. The proposed method for automated determination of centroids and cluster entities in a network of cybersecurity objects allows to increase the speed and objectivity of the analysis of the subject area.

Keywords: cybersecurity, generative artificial intelligence, semantic networks, clustering, centroids, essence of clusters.