2Lbp-RRNS: two-levels RRNS with backpropagation for increased reliability and privacy-preserving of secure multi-clouds data storage

Abstract

Cloud storage as service is the mainstream technology used to retain digital data. However, there are significant risks for confidentiality, integrity, and availability violation associated with the loss of information, denial of access, technical failures, etc. In this article, we propose a two-level 2Lbp-RRNS scheme based on a Redundant Residue Number System with a backpropagation and hamming distance mechanisms for increasing reliability of a configurable and secure multi-cloud data storage. We provide a theoretical analysis of the 2Lbp-RRNS solution as an extension of the classical 2L-RRNS and a variant of fully homomorphic encryption for privacy-preserving, parallel processing, and scalability. We formulate, explain, and prove its main properties to extend existing knowledge within the limits of the critical bounding RRNS assumptions. We show that 2Lbp-RRNS can identify and recover more errors than traditional 2L-RRNS. We provide the upper bounds of the traditional threshold 2L-RRNS and our solution to estimate the number of detectable and correctable errors. We study various data access scenarios and show that it detects 1.58x and corrects 3.37x more errors than 2L-RRNS, on average. We also provide efficient implementations of encoding and decoding algorithms MRC8, and MRC16 based on the Mixed-Radix system, Finite Ring Neuronal Network, and signed binary window method. We evaluate encoding/decoding speeds using three algorithms: Mignotte, MRC8, and MRC16. The experimental system includes seven cloud storages: DropBox, GoogleDrive, OneDrive, Sharefule, Box, Egnyte, and Salesforce. To assess the efficiency of the system on real data, we vary scenarios of the first and second levels. The results show that our solution outperforms MRC8 by 2.53x (1.78x), and Mignotte by 4.83x (11.43x) for the encoding (decoding) speed, respectively