Enhanced Security in Cloud Storage Systems Based on Adaptive Thresholding with Dual Attention Neural Networks

Recent days, cloud storage has revolutionized data management system by providing a flexible, on demand outsourcing solution that eliminated storage management complexities, reduced capital expenditures on hardware and software, which minimized maintenance requirements. Traditional approaches for enhanced security in cloud storage systems had faced significant challenges, which include increased complexity and compatibility issues. Therefore, this paper proposes Adaptive Thresholding-Dual Attention Neural Networks (AT-DANN) which is employed on Benchmarking Evaluation Toolkit for Hypervisor (BETH) dataset. This dataset is preprocessed by using secure Deduplication system, which optimize storage efficiency. Then, content-based chunking and fingerprinting which effectively breaks the data into smaller chunks. After that, proposed AT-DANN is used to adjust threshold values based on data distribution and variability and focused on relevant data features and contexts. This context are hashed by using Multiple Hashing to ensure data consistency and finally, Cross domain authentication is utilized for enabling secure data sharing between different organizations. the proposed AT-DNN for enhanced security in cloud storage systems achieved better results accuracy ratio (89 %), performance ratio (75%), encryption ratio (80 %), efficiency ratio (90 %), evaluation ratio (75 %) when compared with existing Convolutional Neural Networks (CNN).