Cloud Intrusion Detection System

Abstract

Cloud computing is currently reshaping the digital landscape, with a heightened focus on security and privacy concerns for data stored in the cloud. As cyberattacks grow in sophistication and frequency, individuals and organizations alike must prioritize robust intrusion detection systems (IDS). These systems, particularly those utilizing machine learning (ML), excel at identifying network threats but face challenges with large data sizes, leading to decreased performance. Effective feature selection becomes crucial to maintain classification accuracy and prevent information loss. Additionally, addressing imbalanced datasets is vital to mitigate false positives and enhance detection rates. In this study, we propose an enhanced cloud IDS integrating the synthetic minority oversampling technique (SMOTE) for data imbalance and a hybrid feature selection method combining information gain (IG), chi-square (CS), and particle swarm optimization (PSO). Leveraging the random forest (RF) model, our system achieves exceptional accuracies exceeding 98% and 99% on the UNSW-NB15 and Kyoto datasets, respectively. Notably, fewer informative features enhance system efficiency, as evidenced by superior performance compared to existing methodologies