Exploring the Role of Artificial Intelligence in Software Security: A Comprehensive Systematic Review

Abstract

This paper examines the application of Artificial Intelligence (AI) in software security, within a context marked by increasingly sophisticated cyber threats that surpass the limitations of traditional methods, generating an urgent need for more effective solutions. For this purpose, a Systematic Literature Review (SLR) was conducted following the guidelines of Kitchenham [89] and PRISMA, initially retrieving 7,391 documents and refining the corpus to 70 relevant studies published between 2020 and 2025. The analysis focused on examining technologies, theoretical frameworks, and challenges associated with the impact of AI on software security. The results show that the most frequently used techniques are Machine Learning and Deep Learning, with a predominance of algorithms such as SVM, CNN, and Random Forest. In addition, there is a strong concentration of studies in Asian countries, particularly China, and notable development in areas such as Security Integration and Security Enumeration. The findings indicate an ongoing process of consolidation and evolution in the field, although gaps remain, such as the limited attention to emerging approaches like DeepSecAI and the scarce diversification of evaluation criteria. Consequently, future research should prioritize more transparent solutions, the integration of explainable frameworks, and the standardization of metrics that strengthen the comparability and applicability of results.