Updatable encryption in distributed storage systems using key-homomorphic pseudorandom functions

摘要

Distributed storage systems (DSSs) store linear combinations of data across different nodes such that, data lost due to node failures can be restored from surviving nodes. We address key rotation in DSSs: assume that a user wants to remotely change the encryption key of a file stored in a DSS by safely delegating the re-encryption process to the storage network. A problem arises because the data to be re-encrypted is not directly stored in the system (only linear combinations are stored). We propose a solution in which a key-homomorphic pseudorandom function (KH-PRF) is used in counter-mode encryption to make this delegation possible. Our solution requires the homomorphism to be applied several times. Previous constructions of KH-PRFs add noise to their functions which limits the number of times that the homomorphism can be applied. In this paper, we propose the first 'noiseless' key-homomorphic PRFs.