CVALLM: A Cloud Platform Security Assessment Framework Based on Large Language Models

Cloud computing has revolutionized computing and data storage by providing flexible resource management and on-demand services. However, the centralized nature of cloud computing results in significant security issues that pose significant threats to cloud services. The security threat can be alleviated to a certain extent through the cloud platform security assessment. However, the existing cloud platform security assessment framework mainly measures security from the enterprise perspective and lacks methods from the user perspective. For users, much information is internal to the enterprise and cannot be obtained. To address these challenges, we propose a framework called CVALLM to measure cloud platform security from the user’s perspective, which uses a large language model to automatically complete a cloud platform security assessment. First, we utilize a large language model to complete cloud platform information extraction and vulnerability collection from SLA agreements and product description documents that the user can access. Second, to solve the vulnerability assessment problem, we propose a vulnerability severity prediction method based on deep learning. This method considers both the textual description characteristics and the source code characteristics of the vulnerability. Finally, we propose an overall security assessment method for the cloud platform based on a large language model, design quantitative indicators to measure the security of the cloud platform, and automatically generate security reports to facilitate users’ ability to quickly compare and choose the right cloud service.