DDoS Attack Detection Using Deep-Learning Tool with Fused Features and Random Forest Classifier

Recently, a number of attempts and methodologies are created by the attacker to interrupt industry, disturb the significant infrastructures, and online services in order to create financial losses, and reputational damage. The Distributed Denial of Service (DDoS) attack is a common practice which aims to disturb a target server or a network by flooding it with excessive traffic with the help of botnet. When it is detected appropriately, necessary measures can be taken to reduce its impact. This study proposes a Deep-Learning (DL) tool to accurately detect the DDoS. The stages in the proposed scheme includes; data collection and resizing, implementing the chosen DL-scheme to extract the features, identification of the best DL-model for the chosen data and reducing its features using 50% dropout, feature-fusion (FF) and classification with three-fold cross-validation. The proposed DL-tool is verified using the FF with chosen classifiers and this approach help to achieve a DDoS detection accuracy of 100% when the Random Forest (RF) classifier based detection is implemented.