This paper investigates the security features of the distributed archive scheme named Uniformly Random Distributed Archive (URDA). It is a simple, fast and practically secure algorithm that meets the needs of confiden...This paper investigates the security features of the distributed archive scheme named Uniformly Random Distributed Archive (URDA). It is a simple, fast and practically secure algorithm that meets the needs of confidentiality and availability requirements of data. URDA cuts a file archived into fragments, and distributes each fragment into randomly selected n -k + 1 storages out of n storages. As the result, users only need to access at least k storages to recover original file, whereas stolen data from k -1 storages cannot cover original file. Thus, archived files are nothing but sequences of a large number of fixed length fragments. URDA is proved of disappearing both characters and biased bits of original data in archived files, indicating the probabilities of both a fragment and a bit appearing at particular position are uniformly constant respectively. Yet, through running experiments, we found out the risk of likelihood that URDA might be vulnerable to bit pattern attack due to the different ratios of characters appearing in real world files. However, we solved the problem by modifying URDA with variable fragment lengths, which results in that all the bits in revealed sequences are distributed uniformly and independently at random.展开更多
文摘This paper investigates the security features of the distributed archive scheme named Uniformly Random Distributed Archive (URDA). It is a simple, fast and practically secure algorithm that meets the needs of confidentiality and availability requirements of data. URDA cuts a file archived into fragments, and distributes each fragment into randomly selected n -k + 1 storages out of n storages. As the result, users only need to access at least k storages to recover original file, whereas stolen data from k -1 storages cannot cover original file. Thus, archived files are nothing but sequences of a large number of fixed length fragments. URDA is proved of disappearing both characters and biased bits of original data in archived files, indicating the probabilities of both a fragment and a bit appearing at particular position are uniformly constant respectively. Yet, through running experiments, we found out the risk of likelihood that URDA might be vulnerable to bit pattern attack due to the different ratios of characters appearing in real world files. However, we solved the problem by modifying URDA with variable fragment lengths, which results in that all the bits in revealed sequences are distributed uniformly and independently at random.
