CYBERSECURITY FRAMEWORK FOR SECURE CLOUD COMPUTING USING BLOCKCHAIN AND FEDERATED DEEP LEARNING

Abstract

Cloud computing is now a backbone of modern digital services because it offers scalable resources and reduced costs. Yet centralized cloud setups are easy targets for distributed denial-of-service (DDoS) attacks, data breaches, spoofing, insider threats, and malicious traffic. Most existing intrusion detection systems (IDS) train models on a single server, which creates privacy risks, bottlenecks, and poor scalability. This paper proposes a Cybersecurity Framework for Secure Cloud Computing Using Blockchain and Federated Deep Learning (CFSC-BFDL). The framework pairs blockchain-based trust management with a federated convolutional neural network (CNN) so that cloud nodes can detect attacks together without ever sharing raw data. Blockchain handles authentication, tamper-proof logging, and safe model aggregation, while each node trains its own CNN locally and sends only encrypted weight updates to a central aggregator. Experiments on the CICIDS2017 and UNSW-NB15 datasets, validated with 5-fold cross-validation, show an average detection accuracy of 98.4% (+/- 0.31), precision of 97.9%, recall of 97.5%, and F1-score of 97.7%. These numbers beat several existing centralized and blockchain-based IDS approaches. The framework also cut communication overhead by roughly 32% and kept detection latency at 14.2 ms per sample, confirming that combining blockchain with federated deep learning offers a practical path toward stronger cloud security.