Enhancing Cloud Security Using Predictive AI Analysis: A Systematic Review of Literature

Abstract

Abstract: Generally, it has been established that traditional methods of cloud security are plagued with different inadequacies and ineffectiveness. Thus, this study examined how cloud security can be enhanced using predictive AI models. The study adopted the systematic review approach, using the Preferred Items for Systematic Review and Meta-analysis (PRISMA) for data collection. The secondary data were collected from credible databases, which include Google Scholar, Scopus, Taylor and Francis, EBSCOHost, and Emerald, using the appropriate search terms. A total of seventeen (17) articles constitutes the final selected literature. Collected data was analyzed using the “a priori” thematic analysis. The study found that cloud vulnerabilities that are prevalent include detecting anomalies, intrusions, phishing, identify fraud, IoT-enhanced attacks, and malware that compromise infrastructure. Results showed that there are a diverse of predictive AI models used to address these threats include CNNs, Random Forests, SVMs, Naïve Bayes, Decision Trees, LSTMs, BiLSTMs, and Transformers. The findings showed that the predictive AI models used are largely effective in improving cloud security, highlighting how it reduced false positive rates, faster detection speeds, and enhance real-time monitoring performance. Results showed public intrusion datasets such as UNSW-NB15, CIC-IDS2017, and CSE-CIC-IDS2018 are mostly used due to their standardization and structured labelling. Findings showed that there is a heavy reliance on standard classification metrics. Despite all the benefits of AI-enhanced cloud security, challenges such as shortage of high-quality, labelled, and representative datasets affect its effective implementation. The study concludes that predictive AI models enhance cloud security.