Novel Deep Learning Framework for Automated Access Control in Cloud Computing Environments

Abstract

The rapid adoption of Cloud Computing (CC) has significantly increased the complexity of managing secure and efficient access control due to the dynamic nature of cloud environments. Traditional methods, such as static rule-based systems or Machine Learning (ML) models, often fail to adapt to evolving user behavior and the emerging threats. To address these challenges, a novel Malleable Clouded-Leopard tuned Deep Recurrent Neural Network (MCL-DRNN) is designed for automated, real-time access control and resource allocation in CC environments. This framework utilizes cloud access logs, user behavior metrics, device identifiers, geolocation, time-stamps, and environmental variables. Preprocessing includes data cleaning, normalization, and dimensionality reduction using Principal Component Analysis (PCA) to retain critical patterns. The DRNN model is optimized using the MCL method, enabling the system to anticipate complex access patterns and detect deviations from normal behavior. By leveraging the recurrent structure of the DRNN, the system identifies subtle, persistent abnormalities indicative of potential security concerns. Performance evaluations demonstrate that the MCL-DRNN achieves 95% precision, 96% recall, and 95.6% F1-Score, outperforming traditional approaches. The intelligent, adaptive system provides a robust, self-optimizing solution for enhanced cloud security, capable of adjusting to rapidly changing cloud environments and mitigating unwanted access.