Blockchain espoused recalling-enhanced recurrent neural network with fractional discrete meixner moments encryption for cyber security in cloud computing

Abstract

Cybersecurity in cloud computing has become increasingly challenging due to the sophistication of cyber-attacks that target the confidentiality, integrity, and availability of data. Conventional intrusion detection and encryption approaches suffer from high false alarm rates, weak adaptability to emerging threats, heavy computational costs, and limited scalability in real-world deployments. Addressing these limitations, this paper proposes a novel framework that integrates a Recalling-Enhanced Recurrent Neural Network (RERNN) for accurate intrusion classification, Fractional Discrete Meixner Moments Encryption (FDMME) for efficient and robust data protection, and a Direct Acyclic Graph (DAG)-based blockchain for scalable, tamper-proof storage. Preprocessing is carried out using Z-score normalization to handle missing and duplicate values, followed by Entropy–Kurtosis-based feature selection to enhance data quality and reduce dimensionality. Using the NSL-KDD benchmark dataset, the proposed model achieves high detection performance with 98.79% accuracy, 98.78% precision, 98.56% sensitivity, and a Cohen's Kappa value of 98.53%. The false positive rate (0.0144) and false negative rate (0.0025) remain exceptionally low, reducing both false alarms and missed detections. Latency analysis demonstrates near real-time performance with an average processing time of 295 ms, while scalability testing indicates blockchain storage growth from 160 GB to 480 GB and off-chain storage from 15 TB to 45 TB over three years. These findings establish that the proposed RERNN-FDMME-BC-CC framework not only improves classification accuracy and encryption strength but also ensures practical feasibility for deployment in dynamic cloud environments, making it a promising solution for modern cybersecurity challenges.