An Automated Detection of DDoS Attack in Cloud Using Optimized Weighted Fused Features and Hybrid DBN-GRU Architecture

Abstract

Distributed Denial of Service (DDoS) attack is a severe type of attack. It affects the cloud in terms of loss of repudiation, service outage, financial loss, and data loss. Hence, the optimal feature selection and the hybrid learning-based classifier are developed for evading the normal function disruption in the cloud server. After pre-processing, the resultant data is subjected to the ensemble feature selection. The first feature is selected optimally by the Probability of Fitness-based Billiards-Inspired Optimization (PF-BIO) algorithm, the second feature selection is acquired through the Fisher discriminant, and the last feature selection is carried out by feature correlation with class. Consequently, the resultant of these feature selections is concatenated with the weight parameter to provide weighted fused features, where the weight is getting optimized by the PF-BIO algorithm. Finally, a newly developed hybrid-based learning model, where the Deep Belief Network with the Gated Recurrent Unit (DBN-GRU), in which the hyper parameters are tuned and optimized by the PF-BIO algorithm. Thus, the performance is compared with the existing approaches. Throughout the result analysis, the accuracy of the designed method is 97.05%. Hence, the proposed model proves the efficiency for the detection of DDoS attacks precisely.