Topic recognition and refined evolution path analysis of literature in the field of cybersecurity

• Published in: PloS one Vol. 20; no. 2; p. e0319201
• Main Authors: Zhu, Yanfeng, Li, Zheng, Li, Tianyi, Jiang, Lei
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 21.02.2025; Public Library of Science (PLoS)
• Subjects: Bibliometrics; Comparative analysis; Computer Security; Cybersecurity; Data mining; Data security; Datasets; Evaluation; Evolution; Humans; Methods; Natural language processing; Path analysis; Recognition; Scientometrics; Social science research; Trends; China
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0319201

Using text analysis techniques to identify the research topics of the literature in the field of cybersecurity allows us to sort out the evolution of their research topics and reveal their evolution trends. The paper takes the literature from the Web of Science in the field of cybersecurity research from 2003 to 2022 as its research subject, dividing it into ten stages. It then integrates LDA and Word2vec methods for topic recognition and topic evolution analysis. The combined LDA2vec model can better reflect the correlation and evolution patterns between adjacent stage topics, thereby accurately identifying topic features and constructing topic evolution paths. Furthermore, to comprehensively evaluate the effectiveness of the LDA model in topic evolution analysis, this paper introduces the Dynamic Topic Model (DTM) for comparative analysis. The results indicate that the LDA model demonstrates higher applicability and clarity in topic extraction and evolution path depiction. In the aspect of topic content evolution, research topics within the field of cybersecurity exhibit characteristics of complexity and diversity, with some topics even displaying notable instances of backtracking. Meanwhile, within the realm of cybersecurity, there exists a dynamic equilibrium between technological developments and security threats.