Improved restricted Boltzmann machine-based optimization model for the network security system in cloud environment

Abstract

This paper introduces a novel approach to enhancing computer network security through deep learning and cloud technologies, focusing on combating insider DDoS attacks in cloud computing. By employing the IRBM architecture and a trust-oriented secure cloud framework, the work utilizes parameter tuning in RBM through the Gradient-Based Optimization (GBO) algorithm resulting in the proposed IRBM-GBO model. This model significantly outperforms SVM, SC-CNSP-ML, and RBM in terms of latency and Packet Delivery Ratio (PDR), showing 21.67%, 14.55%, and 9.62% improvements in latency, and 10.23%, 5.43%, and 12.79% in PDR, respectively. The research highlights the potential of combining soft computing with deep learning and cloud technologies to identify cyberattacks by emphasising improvements in latency, Quality of Service (QoS), throughput, and PDR across various cybersecurity datasets.