Cloud Privacy Preservation Using Improved Squeezenet Based Data Sanitization and Improved Lyrebird  Optimization-Based Optimal Key Generation

Abstract

Maintaining privacy in the cloud is critical, which implies reliable and effective models that are adapted to the difficulties presented by cloud settings. This paper introduces a comprehensive privacy preservation model tailored specifically for cloud environments, comprising five important phases such as data acquisition, normalization, feature extraction, sanitization and restoration. It begins with the meticulous collection of data from diverse sources, followed by a normalization process to standardize and cleanse the acquired data, ensuring uniformity and consistency. Subsequently, crucial features such as improved entropy and statistical measures are extracted from the normalized data to provide valuable insights. The pivotal data sanitization phase employs three key processes: optimal key generation using the Improved Lyrebird Optimization Algorithm (Imp-LOA), key tuning through deep learning with the Improved SqueezeNet, and Kronecker product operation to determine the encryption process. On the other end, the data restoration process is done, which is the reverse process of sanitization, to retrieve the data. The proposed model addresses the optimization objectives including hiding failure, data preservation ratio, modification degree, and privacy. The Improved Lyrebird Optimization Algorithm generates encryption keys based on the natural behavior of lyrebirds, providing superior safety. Anticipated outcomes of this research encompass MATLAB-based simulation and investigational analysis, benchmarking against existing methods to evaluate the model’s efficacy in terms of security, time efficiency, and other pertinent metrics. Through this comprehensive analysis, the proposed model’s superiority in safeguarding privacy in cloud environments has been demonstrated, marking a significant advancement in privacy-preserving techniques within cloud computing.