Machine Learning Based Approach for Handling Imbalanced Data for Intrusion Detection in the Cloud Environment

Cloud computing is a flexible on-demand service provider for pooled access to various software resources. Due to the high demand for the cloud for data storage and access, there are chances of getting it vulnerable to threats associated with data stealing and modifications to it, which results in poor management of cloud environments. Thus, detecting intrusions into cloud platforms is essential to providing better services to users and saving their important data and resources from being misused. Machine learning (ML) plays a massive role in detecting intrusions in cloud environments. Still, the most common problem that is faced is data imbalance as the data collected will consist of few intrusion samples. In contrast, the maximum will be normal ones, which makes the ML biased towards the majority class. Hence, to provide a solution to this problem, this paper aims to propose an approach that utilizes different ML techniques without using any over-sampling or under-sampling techniques for balancing the data. The different ML methods on the imbalanced data are trained to find the most optimal ML method by just optimizing the hyperparameters. The highest accuracy of 100% is achieved by three models, i.e., decision tree, random forest (RF), and XGBoost. However, when comparing the training time, DT and RF are the most promising algorithms for detecting intrusions on big imbalanced data without using any data balancing technique.