Model for Describing Processes of AI Systems Vulnerabilities Collection and Analysis using Big Data Tools

To assess the cybersecurity of attacked Artificial Intelligence (AI) systems, it is very important to have trustworthy, up-to-date and systematic information about vulnerabilities. The paper provides a critical analysis of existing methods and tools for collecting information about vulnerabilities to assess the cybersecurity of AI systems, and defines indicators of the quality of this information, primarily completeness and reliability. The processes of collecting and analyzing AI systems vulnerabilities are described by a multi-level IDEF (Integrated Definition) model. It consists of the following stages: systematization of information sources, its processing by groups of sources and coordination using Big Data tools; selection and presentation in a certain format of information about vulnerabilities of the studied AI system; assessing the criticality of system vulnerabilities using various methods, in particular, IMECA (Intrusion Modes and Effects Criticality Analysis). The features of detailing the model of the Natural Language Processing (NLP) as a Service system are illustrated. The directions of future research and the development of tools for implementing the proposed model for collecting and analyzing vulnerabilities of AI systems are substantiated.