Intelligent test case generation from textual security requirements in SCRUM: an NLP-driven approach

This paper presents a method for automatically generating security-oriented test cases from textual requirements in SCRUM environments using Natural Language Processing. The proposed approach has combined transformer-based semantic analysis with behavior-driven development test templates to extract and translate functional, non-functional, and misuse-case security requirements. The solution has been tested on 30 real-world requirements derived from agile software projects. Evaluation results have demonstrated that the system achieved 91% precision, 93% recall, and complete (100%) coverage of input requirements. Compared to manual testing, the method has reduced the time required for test design by approximately 78% and revealed 65% more critical security vulnerabilities. The generated test cases have been structured to support integration with behavior-driven development and continuous integration/continuous deployment workflows. Overall, the results indicate that automation based on Natural Language Processing can substantially enhance the quality and efficiency of security validation processes within agile development environments.