Deep Learning Method for Detecting and Mitigating Distributed Denial of Service Attacks with Imbalanced Data

A Distributed Denial-of-Service (DDoS) attack is a malicious attempt to reduce normal traffic service, or network by spreading a target or its associated infrastructure with an internet traffic flood. The salient features of Cloud Computing (CC) are being leveraged by attackers for launching severe DDoS attacks. However, DDoS attacks are very challenging to detect and mitigate the attacks accurately due to their distributed nature. This research proposes the Deep Convolutional Neural Network-based Support Vector Machine Activation Kernel (DCNN-based SVMAK) for detecting and mitigating DDoS attacks accurately using Deep Learning (DL). Initially, the CIC-DDoS2019 and Bot-IoT datasets are utilized to evaluate the proposed technique, and min-max normalization is employed for normalizing the obtained data. Then, the data imbalance issue will be overcome using the Weighted Optimized Synthetic Minority Over-Sampling Technique (WSMOTE). The Hybrid Golden-Seagull optimization (HGSO) is performed by changing sin into cos function for feature selection. The DCNN-based SVMAK is established for the detection and classification of DDoS attacks. If the attack is detected, the BAIT approach is used to mitigate the DDoS. The existing approaches like CNN-Bidirectional Long Short-Term Memory (CNN-BiLSTM), AutoEncoder-Multi-layer Perceptron Network (AE-MLP), Wrapper Feature selection-based Hybrid DL (WF-HDL) are used to compare with the proposed approach. The proposed approach achieves better accuracy of 99.81% and 99.87% for CIC-DDoS2019 and Bot-IoT datasets compared to CNN-BiLSTM, AE-MLP, and WF-HDL respectively.