GUARDIANS OF THE DATA GALAXY: A FEDERATED AI AND CLOUD SYNERGY FOR ZERO-TRUST CYBERSECURITY MODELS
Habiba Mahamad · International Journal of Education Humanities and Social Science · 2024
As cyber-attacks evolve, traditional centralized security paradigms struggle to offer data privacy, scalability, and real-time threat detection across distributed systems. This paper introduces a novel solution that combines Federated Artificial Intelligence with Zero-Trust security principles to create a secure, decentralized cybersecurity system. The proposed methodology integrates a Network Intrusion dataset from Kaggle with real-time and synthetic data collected from enterprise networks, IoT devices, and cloud infrastructures. With the simulation of an actual attack surface, this setting allows for in-depth training and testing of intrusion detection models. Data preprocessing involves normalization techniques that normalize features at nodes such that training is performed efficiently and uniformly in the federated environment. A Federated Convolutional Neural Network is deployed on decentralized edge nodes where local training is performed by each node without compromising data privacy. The model updates are securely aggregated by a central cloud server, without raw data transfer and reducing the risk of exposure. The blend of Zero-Trust paradigms and federated learning provides real-time decision-making capabilities with privacy and system integrity. Experimental outcomes reflect the supremacy of the suggested framework over traditional centralized and basic federated models regarding accuracy, rate of convergence, and anomaly detection accuracy. Precision-recall plots and training loss confirm improved model stability and lower false positives. The framework was implemented using Python and tested on benchmark datasets, and the accuracy was 94.3%, substantiating its applicability in safeguarding distributed and dynamic network systems. The system paves the way for privacyaware future intelligent cybersecurity systems.