International Journal of Science, Engineering and TechnologyAuthors: Ashwini Gulhane, Abdul Raafeh, Mohammed Affanuddin, Ma Khizer Moinuddin
Abstract: The transformation towards the adoption of cloud computing has provided modern enterprises with the major benefits of easily extending their resources and being able to use them in various ways without any restrictions, but at the same time it has brought the enterprises a new set of complicated and changing security threats that are very hard to deal with. The conventional security countermeasures which are primarily based on static rule-based mechanisms, perimeter defenses, and Multi-Factor Authentication (MFA), have shown to be of limited effectiveness against the advanced attack vectors of insider threats, privilege escalation, and especially the large-scale, rapidly changing Denial-of-Service (DoS) and Distributed Denial-of-Service (DDoS) campaigns.[1, 1] These traditional systems have been rendered ineffective most of the time when it comes to detecting zero-day exploits and sophisticated lateral movement which is characteristic of modern botnet operations. In this paper, we describe the process of creating, developing, and assessing SmartTrust, an advanced hybrid deep learning framework that is specifically designed to carry out real-time threat detection in cloud environments while being totally in accordance with Zero-Trust Architecture (ZTA) principles.[1, 1] SmartTrust is a system that is built on a composite deep learning core, which is the result of the integration of Convolutional Neural Networks (CNN) for the analysis of spatial patterns, Long Short-Term Memory (LSTM) networks for the understanding of temporal dependencies and, lastly, Transformer models for the extraction of global contextual relationships in network traffic and user behavior logs. A pivotal feature of the framework is its explicit optimization layer, realized through Reinforcement Learning (RL), which allows for adaptive decision-making and continuous policy adjustment based on real-time contextual signals; thus, Concept Drift is dynamically countered. Besides, in order to maintain unbroken forensic integrity and also compliance alignment with ZTA, the system implements tamper-proof Blockchain-Based Logging for all
