Abstract

Due to such rapid advancement in the field of cloud computing, the need has drastically emerged for much-improved security protocols in order to save it from nefarious cyber attacks. This would also entail understanding that, in such a dynamic setting, the role of Intrusion Detection Systems is imperative, acting as a sort of support system that helps in identifying malicious activities and intrusions. The paper explains the discussion of various machine learning techniques employed for intrusion detection in cloud systems. We test a wide range of models that use Random Forest, XGBoost, DNN, Decision Trees, KNN, Logistic Regression, SVM, and Naive Bayes to check their efficiencies for identifying various forms of assaults. This research paper also defines the respective pros and cons for each one of them. The following analysis makes its conclusion based on critical performance factors such as accuracy, recall, and F1-Score, which would be fundamental in determining the efficiency of such models in handling the forms of current network attacks. It is clear from the results that ensemble techniques such as XGBoost and Random Forest models are superior to other models evaluated. These models effectively balance the trade-off between high detection accuracy and low false positive rates, making them the most robust choices for multi-class, complex cloud security systems.