EFS-CCE: An Ensemble Feature Selection using Combined Criteria Evaluation for Scalable and Reliable DDoS Detection in Cloud Computing Environments

The Cloud computing Environment (CCE) has recently become an integral part of enterprises and an effective platform for data storage and processing. As its use continues to expand, so does the exposure to application-level distributed denial-of-service (DDoS) attacks, with HTTP-GET attacks posing a notable threat. These types of attacks rely on CCE infrastructure for service retrieval and resource consumption. An effective intrusion detection mechanism for this purpose requires Feature Selection (FS) methods to overcome redundancy and improve classification performance at a low computational cost. This paper introduces an Ensemble Feature Selection (EFS) approach by integrating the results of Information Gain (IG), Gain Ratio (GainR), the Chi-squared statistic, and the Symmetrical Uncertainty measure using an efficient Majority Voting Mechanism (MVM), followed by further optimization of the selected attributes using Principal Component Analysis (PCA). Comprehensive experiments are conducted on a standard HTTP-GET log CCE dataset for classification using Machine 1 Learning (ML) and Deep Learning (DL) classifiers, achieving promising state-of-the-art performance with an accuracy of 99.8269%, a precision of 0.996%, and a dimensionality reduction from 26 to 15 features without loss of performance.