An Efficient DDoS Attack Detection Using Chaos Henry Gas Solubility Optimization Weight Initialization Based Rectified Linear Unit

Abstract

The Denial of Service (DoS) attacks is one of the main issues faced by cloud service providers due to their intricate nature. The main aim of this attack is to disrupt the services of authorized users by forwarding massive malicious requests to the victim system. Even though the modern Artificial Intelligence-powered intrusion detection system offers improved benefits, it suffers from analyzing the traces of intrusion in the massive network flow. This article presents a novel Chaos-based Henry Gas Solubility Optimization—Weight initialization based-rectified linear Unit (HGSO-WIB-ReLU) framework to identify different types of DoS attacks like HTTP flood attacks, DNS flood attacks, and ICMP flood attacks. The main aim of this technique is to identify the low rate of DoS attacks which the state-of-art techniques often based on statistical analysis and machine learning are incapable of accomplishing. Convolutional neural network architecture incorporates the WIB-ReLU activation function primarily to prevent vanishing gradient issues and provide efficient training. The WIB-ReLU framework’s hyperparameters are primarily improved using the HGSO method by increasing the classifier’s accuracy rate for detecting DoS attacks. In this way, the proposed model minimized the high false positive rate in identifying low-rate DoS attacks in large datasets. Different benchmark datasets such as BUET- DDoS2020, CIC DoS Attacks, and Low Rate DDoS are used to conduct the experiments and the efficiency of the model is verified in terms of accuracy, F1-measure, and precision score. When evaluated using the BUET-DDoS, CIC-DoS, and Low rate DDoS datasets, the proposed model offers an accuracy of 97%, 96.67%, and 96%, respectively. The experimental analysis conducted demonstrates the effectiveness of the proposed HGSO-WIB-ReLU framework when compared to the different state-of-art methodologies.