Secure Cloud Storage with Joint Deduplication and Erasure Protection

Abstract

This work proposes a novel design for secure cloud storage systems using a third party to meet three seemingly opposing demands: reduce storage requirements on the cloud, protect against erasures (data loss), and maintain confidentiality of the data. More specifically, we achieve storage cost reductions using data deduplication without requiring system users to trust that the cloud operates honestly. We analyze the security of our scheme against honest-but-curious and covert adversaries that may collude with multiple parties and show that no novel sensitive information can be inferred, assuming random oracles and a high min-entropy data source. We also provide a mathematical analysis to characterize its potential for compression given the popularity of individual chunks of data and its overall erasure protection capabilities. In fact, we show that the storage cost of our scheme for a chunk with r replicas is O(log(r)/r), while deduplication without security or reliability considerations is O(1/r), i.e., our added cost for providing reliability and security is only O(log(r)). We provide a proof of concept implementation to simulate performance and verify our analytical results.