An Intelligent Retrieval System for Similar Information System Vulnerabilities Based on Knowledge Graph

Junzhe Dong, Shuo Geng, Xiong Zhang

Abstract: With the continuous improvement of information technology, information systems have been widely used in the operational processes in various organizations. While information systems provide convenience, information system vulnerabilities also threaten the privacy and security of users. When vulnerabilities occur, quickly resolving them can minimize the losses for organizations and users. At present, the determination of the types and levels of vulnerabilities in information systems is based on manual work, which requires a high level of expertise and a long time for personnel involved in resolving vulnerabilities. Automatic classification and hazard level assessment of information system vulnerabilities can help vulnerability solvers quickly evaluate vulnerabilities and determine the appropriate methods and order of vulnerability resolution based on vulnerability categories and hazard levels. Therefore, this article constructs an information system vulnerability classification model and an information system vulnerability hazard level evaluation model based on the BERT model, thereby automating this process. In addition, based on classification and hazard level prediction models, retrieving similar vulnerabilities intelligently can serve as references for vulnerability solvers, which can improve the efficiency of vulnerability resolution. Therefore, this article proposes to implement an intelligent retrieval system for similar vulnerabilities. Firstly, the BERT model is used to classify vulnerabilities and evaluate their hazard levels. Subsequently, this study constructs a knowledge graph for information system vulnerabilities using semi-structured information system vulnerability data. By combining the vulnerability classification model, the hazard level assessment model, and the information system vulnerability knowledge graphs, this study realizes an intelligent retrieval system for similar vulnerabilities.