Fortifying Cloud Security with a Single Candidate Optimized Deep Spectral Convolution Network for Privacy Safeguarding in Trusted Environments

Abstract

Cloud computing has been widely adopted across diverse technological domains as an internetbased, self-service platform for delivering computing resources, transforming infrastructure and technology management. However, the increasing migration of sensitive data to cloud environments has intensified concerns regarding security and privacy. Cloud service providers face increasing threats from malicious users, intrusions, breaches of confidential information, information disclosure, and advanced cyber-attacks. Building security mechanisms for cloud computing is also complicated due to the distributed nature of the cloud infrastructure. Moreover, the common intrusion detection techniques usually involve higher computation costs, lower generalizability, and poorer performance against the evolving threats. In light of the above restrictions, this paper introduces a Single Candidate Optimized Deep Spectral Convolutional Network (SiCO: DeS-CNN) for cloud intrusion detection and privacy-aware security enhancement. It combines Composite Maximum Likelihood Code-Length (CompMLCL) preprocessing, Minimal Memory Attention Loop Transformer (MiMeALT) feature extraction, Deep Separable Convolutional Neural Networks (DeSCNN), and Single Candidate Optimization (SiCO) for increased detection accuracy and computational efficiency. The effectiveness of the proposed technique is shown on the NSL-KDD and UNSW-NB15 datasets in achieving an accuracy of 98.9% and precision and recall of 97.5% and 96.8% respectively, while reducing computation latency and energy efficiency by 90% and 92%, respectively compared to the existing techniques.