Ensemble Feature Selection Method and Combining Classifier Approach for DDoS Detection in Cloud

Abstract

            Through the substantial growth of cloud computing, exchanging critical computing resources online is generating new business models. Distributed Denial of Service ( DDoS ) attacks threaten-cloud infrastructure. Cost-effective and company losses are the direct ripple impact of service failure, which can be offset by identifying DDoS attacks early. This paper introduces a tree-based classification model combination with an Ensemble-based Feature Selection Framework (EBFM) inthe process of detecting DDoS attacks. The proposed methodology tested the NSL-KDD dataset. EBFM selects 12 key features from the dataset and the detection accuracy was achieved by integrating random tree and NBTree algorithms using the scheme named as sum rule to 99.82%, which increases the accuracy rate to 99.79% compared to the other approaches. The proposed method, like other classification methods, requires high detection rates and accuracy.