Data security technologies for robust cloud storage services.

摘要

Cloud storage has been recognized as one of the most potentially profitable cloud computing applications. However, potential users would be wise to question whether the integrity, confidentiality, and availability of their data are not violated in cloud storage. This is not only due to the fact that research in cloud computing security is not mature, but also because negative experiences have shown that none of the large-scale storage services is completely reliable and any of them may lose or corrupt customer data. This dissertation focuses on improving data security for cloud storage from three different perspectives, namely, integrity, confidentiality, and availability.;The main contribution to improve integrity is to address the potential disputation problem. A consistent solution is proposed based on two novel non-repudiation protocols. One is a two-party non-repudiation protocol (TPNR) and the other is a multi-party non-repudiation protocol (MPNR). The protocols solve the disputation problem by requiring that certain evidence is generated and held by all the parties involved in a data transmission. The MPNR protocols can also solve additional problems including protecting against Collusion attacks and Roll-Back attacks.;The primary approach for improving confidentiality is a mechanism called D-DOG (Data Division and Out-of-order keystream Generation) for a distributed storage network. The D-DOG creates cipher blocks by dividing the plaintext data into multiple blocks and encrypting them, such that the keystream is generated by abstracting bits from the data blocks in a pseudo-random out-of-order manner.;My work on availability lies in two improved XOR based decoding scheme called EDP (Enhanced RDP code) and EEO (Extension of the EVENODD code). In theory when implemented in parallel, the decoding velocity of the EDP scheme can be improved by about 40% without changing any disk configuration for the RDP code. EEO is an efficient array-like scheme which utilizes optimal decoding at a cost of adding two parity symbols on the parity devices. The decoding velocity of the EEO scheme is 100% faster than EVENODD and about 267% faster than the RDP code in parallel mode at the cost of two extra symbols.