Mitigating data leaks and optimizing task scheduling in cloud computing using EfficientNetV2 with federated learning and blockchain

Mitigating data leaks and optimizing task scheduling in cloud computing using EfficientNetV2 with federated learning and blockchain

연구 분야: Software Development

학회: Cluster Computing

Cloud computing offers scalable, on-demand services but faces critical challenges related to security vulnerabilities, data leakage, and inefficient task scheduling. Existing solutions often fail to provide a unified approach that ensures both security and optimal resource utilization. This study proposes a novel Secure and Optimized Federated Cloud Framework (SOFCF) that integrates EfficientNetV2-based Federated Learning (FL), Kubernetes-based task scheduling, and blockchain-enhanced security to address these issues. The proposed model prevents data leakage by employing FL for decentralized model training, ensuring data remains local while contributing to a global model. EfficientNetV2 accelerates training convergence, improving classification performance. Kubernetes dynamically schedules tasks based on real-time demands, utilizing a Multi-Armed Bandit (MAB) fine-grained resource allocation strategy to optimize CPU and memory usage. EfficientNetV2 accelerates training convergence, achieving 99.13% classification accuracy on the CVC3345 dataset, outperforming existing models. The proposed model improves resource utilization by 19%, reduces data leakage incidents by 85%, and reduces makespan by 28%. The proposed model reduces response time significantly, achieving 155 ms, showcasing its efficient task scheduling and resource management. This research establishes a trustworthy, decentralized, and efficient cloud computing model, ensuring privacy-preserved and optimized resource management in distributed environments.