Optimal Generative Adversarial Network-Bidirectional Long Short-Term Memory-Based Byzantine Failure Prediction and Failure Tolerance in Cloud Computing

The research leverages the Optimal Generative Adversarial Network (OGAN)-Bidirectional Long Short-Term Memory (BiLSTM) architecture towards present a novel framework for the prediction and tolerance of Byzantine failures in cloud computing. The input dataset, derived from the KDD dataset, undergoes meticulous preprocessing to eliminate redundant characteristics and outliers, employing the Min-max normalization approach. The preprocessed data can be then integrated into the Byzantine Failure Prediction Framework (BFPF) within the cloud environment. The OGAN-BiLSTM model, enhanced by the Improved Dwarf Mongoose Optimization (IDMO) Algorithm, is utilized aimed at optimal hyperparameter selection, significantly improving classifier performance, Byzantine Failure Prediction, and tolerance. For improved storage security, a two-way encryption scheme, as Modified Homomorphic Encryption Algorithm (MHEA) for secure non-intrusive data, and subsequently stored in a blockchain (BC) based cloud environment, is implemented. The innovation of the work lies in its unique combination of progressive technologies to improve cloud computing (CC). The suggested model's presentation is assessed utilizing a variety of measures, like sensitivity, recall, f-measure, specificity, accuracy, precision, and execution time. The projected paradigm will be incorporated into the Java development environment. The accuracy of the suggested model for CC Byzantine failure prediction and tolerance is 98.69%. The suggested approach improves data security in cloud environments and greatly increases the accuracy of failure predictions.