Enabling dynamic proof of retrievability in regenerating-coding-based cloud storage

Abstract

To protect outsourced data in practical cloud storage against corruptions, enabling the integration of dynamic proof of retrievability (DPoR) and efficient recovery for cloud storage becomes significant and critical. However, in general PoR cloud storage, as all or part of data files are encoded, frequent or small updates require the updates of all related (encoded) file. Thus, this work studies the problem of constructing a novel regenerating code to be compatible with dynamic provable data possession (DPDP) into a DPoR system. Specifically, a class of exact-regenerating systematic code is presented to efficiently combine DPDP using the product matrix framework. In addition, a practical DPDP scheme is proposed and implemented for the specific regenerating code, while preserving the combined properties of default data integrity protection, efficient dynamic data updating, fault tolerance and repair traffic saving. Our DPDP scheme is based on the new Memory Adversary model specifically brought by dynamic operations. It allows different parameters to be fine-tuned for the performance-security tradeoff. We implement and evaluate the overhead of our DPDP scheme in cloud storage under different parameter choices. We demonstrate that DPDP can be feasibly integrated into our specific regenerating code.