A Bio-Inspired Feature Selection and Ensemble Classification for DDoS Detection in Cloud

Abstract

The investigation of cloud computing is becoming more popular in both the business world and the academic world. The use of cloud computing presents many opportunities for growth and improvement for cloud service providers as well as end users. Due to the dramatic increase in demand for cloud computing, data security has emerged as a primary area of concern. There have been a great deal of risks that make the use of cloud computing more difficult. Detecting distributed denial of service attacks is a key bottleneck in the cloud technology industry. The development of an efficient attack detection technique is a challenging endeavor because of the intricate interactions between nonlinear interruption activities, aberrant system traffic behavior, and other variables. As a result, establishing preventive solutions against these threats is critical for the broad adoption of cloud computing. This work presented a combination of the bio-inspired feature-choosing method Particle Swarm Optimization (PSO) with the classification methods Logistic Regression, Gaussian, and Random Forest as an ensemble technique for Distributed Denial of Service (DDoS) attack detection. The Bio-Inspired Feature-Selection and Ensemble-Classification DDoS-Detection (BIFSED) output is finalized by combining the results of each categorization technique. To determine the final DDoS classification, we employed a certain threshold and a vote of simple majority. The performance results with the NSL-Knowledge Discovery-Dataset (NSL-KDD) dataset showed that the BIFSED approach, with thirteen characteristics and ensemble techniques, outperforms a complete set of features and different classification methods in the literature using logistic regression, Gaussian, and random forest classification methods.