Research On Data Security And Coding Techniques For Distributed Storage Systems

With the rapid development and increasing improvement of distributed networking technology such as cloud computing and big data,secure handling,storage and transmission for vast amounts of data,currently have become a global issue.Towards three basic data security requirements,i.e.,integrity,reliability and confidentiality,researchers in the distributed storage architecture carry out the in-depth investigation,and have achieved many important fruits in the new auditing,coding and encrypting technology.Nevertheless,existing achievements are still facing some challenges in the aspects of security,theory and practice,such as security assumptions of public auditing,potential threats of dynamic operation,privacy protection of storage data,and theoretical exploring of new coding etc.Henceforth,how to deal with the security threats faced by the distributed data storage,providing appropriate mechanisms for guaranteeing data security,and deeply analyzing and improving the existing secure data storage model,are our main research content.In this article,by centering around the major security challenges,we conduct the research on data security and coding technique for distributed storage systems.Based on the summary and analysis of existing technical solutions as well as theoretical results,we deeply investigate and carry out researches on information-feedback-based audit programs,secure storage of dynamic updating data,privacy protection of new coding,and theoretical analysis of some new codes etc.Specifically,our results are as follows:1.We present a data integrity protection scheme with untrusted third parties.Existing public audit schemes always assume that third party auditor is trusted or semi-trusted.However,in the real scenario of data integrity checking,third party auditor may be malicious,which would directly affect the final checking result.To this end,we propose an audit model with secure computing proxy based on information feedback.By transferring the major computing tasks to a third party,users can achieve their own simple and efficient completion of the final audit tasks.2.We propose a secure storage model for dynamic updating data based on regenerating codes.Regarding the real data storage platform,data often needs dynamic operations like deletion,adding and updating etc.Due to the redundant structure of storing data,dynamic operations bound to affect all of the relevant data blocks.In addition,if the attacker targets destructing the dynamic changeable data blocks,it will inevitably lead to the permanent unrecovery of all related data blocks.Therefore,we consider the dynamic proof of retrievability scheme based on systematic regenerating codes.Using the random permutation and batch updating technology,we ensure that the data can complete the secure efficient dynamic operations in a high probability.3.We characterize the secrecy capacity of MSR-codes-based distributed storage systems.As the optimal regenerating codes with minimum storage,MSR codes have a wide range of practical applications.However,since the storage capacity of one single node for any MSR code is strictly less than the corresponding repair bandwidth,the characterization of secrecy capacity under the extended eavesdropping model was once considered as an open problem.As for it,we consider from the perspective of information theory,combine with the relevant knowledge of finite field,and deeply study the inherent information-theoretic features in the distributed storage architecture.Finally,we give the specific characterization of their secrecy capacity.4.We analyze the data secrecy of MSCR-codes-based storage systems.MSCR codes are minimum storage cooperative regenerating codes that can simultaneously and effectively repair multiple failed nodes,and thus are adaptable to most practical applications of distributed storage.Also in the extended eavesdropping model,the design of MSCR codes with privacy protection,will be of important practical value.However,for MSCR codes,repair data may vary with different repair groups or help nodes sets,which definitely results in more information leakage.To this end,we introduce a new concept that is “stable” MSCR codes.By analyzing two existing kinds of MSCR codes,we find that they both are unstable.We prove that one class does not have any data secrecy capacity,while the other can be converted to a stable MSCR code.By use of information theory,we further characterize the secrecy capacity of stable MSCR codes under certain situations.5.We improve the upper bounds of systematic length for linear systematic MSR codes.For practical applications,in view of computational complexity and data access cost,linear systematic MSR codes are most preferred.However,due to MSR codes' inherent features as well as additional linear and systematic properties,there exists natural restriction within three parameters,i.e.,single-node storage capacity,number of parity nodes and systematic length.In response to this open problem,we follow and consider the principle of interference alignment as well as the technology of linear subspace and linear operator.By verifying the linear independence of the original coding matrices and constructing new independent matrices,we give the improved upper bounds of systematic length,consequently reinforcing the understanding of the construction principle for linear systematic MSR codes.Based on the analysis of the data security requirements for distributed storage systems,the above results put forward defensive solutions against the potential hazards of the prior technical schemes,and fully answer couple of important problems in distributed storage coding theory.In this regard,they are of important theoretical and practical value for promoting the secure applications of distributed storage systems.