Threat Analysis Using Topic Models in Large-Scale Vulnerability Databases and Security Incident Case Documents

• Published in: 2021 IEEE International Symposium on Technologies for Homeland Security (HST) pp. 1 - 6
• Main Authors: Koyanagi, Hiroki, Takaragi, Kazuo, Wohlgemuth, Sven, Umezawa, Katsuyuki
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08-11-2021
• Subjects: Analytical models; Data models; Natural language processing; Natural languages; Security; Topic model; Vulnerability Information
• DOI: 10.1109/HST53381.2021.9619846

It is crucial to design products bearing security in mind from the initial development stage. Consequently, many threat analysis support tools have been developed. However, it is difficult to determine the inherent threats in various designed documents written in natural language, which is used in the initial development stage. It is not uncommon to find attacks that closely resemble past attacks. In addition, many designs are limited in the number of data they can handle. We propose a method of extracting existing vulnerabilities similar to those used in the attack by collating a large vulnerability database with existing attack cases using Latent Dirichlet Allocation, one of the topic model methods. We apply the proposed method to several cases and verify its effectiveness.