EXPLORING MACHINE LEARNING IN SECURING SOFTWARE REQUIREMENT SPECIFICATIONS: A SYSTEMATIC LITERATURE REVIEW

Abstract

Securing Software Requirement Specifications (SRS) is a critical step in ensuring robust and secure software systems, as vulnerabilities at this stage can propagate throughout the software development lifecycle. This systematic literature review explores the application of machine learning (ML) techniques in enhancing SRS security, focusing on identifying, mitigating, and predicting vulnerabilities and inconsistencies in requirement specifications. The study synthesizes recent advancements in the field, including the use of natural language processing (NLP) for extracting and analysing security requirements, predictive models for vulnerability detection, and hybrid approaches combining traditional security frameworks with ML. Key findings highlight the potential of ML to automate and improve the accuracy of security analysis in SRS while addressing challenges such as data scarcity, model interpretability, and domain-specific complexities. By identifying gaps in existing research, this review provides insights into emerging trends and proposes directions for future studies to advance the integration of ML in securing SRS. This work contributes to bridging the gap between academic advancements and industrial practices, paving the way for more secure software development processes.