Machine Learning for Cloud Security: A Systematic Review

Abstract

The research investigates cloud security by executing machine learning (ML) applications to obstruct cyber threats within cloud infrastructure. This research uses different ML approaches to construct IDSs by testing multiple IDS variations using the NSL-KDD attack simulation dataset pool. The evaluation shows J48 and PART having 88% accuracy while Decision Stump reaches 90% accuracy and Decision Table reaches 93% and AODE reaches 99% accuracy. The ability of Average One Dependency Estimator to detect patterns in lengthy multi-dimensional datasets makes it the optimal solution. Feature selection techniques including Intrinsic Information together with Information Gain and Gain Ratio improve both data performance and operational efficiency by reducing dimensions. The conversion of numeric features to nominal types ensures maximal algorithm performance because several ML models function more efficiently with categorical data. The system undergoes thorough evaluation to determine its accuracy levels and speed performance and cloud application compatibility. The research provides an effective security platform which merges state-of-the-art ML approaches with elaborate preprocessors and evaluators to enhance security protection against progressively changing threats.