Today, cloud technology continues to evolve. Using cloud services allows you to get financial benefitsю Still, many companies and users are in no hurry to transfer their infrastructure to the cloud due to incompletely resolved problems related to the security of data storage and processing. Since in the case of storing and processing open data, a cloud provider gets access to the data, and an attacker can also gain by hacking an account. In the case of encrypted data transmission to the cloud, confidentiality will be preserved only in data storage tasks, since data processing will require a decryption task. The use of homomorphic ciphers allows the processing of encrypted data without violating their privacy. Homomorphic encryption is actively beginning to be used in machine learning tasks to transfer and ensure the confidentiality of resource-intensive operations for training a neural network in the cloud. The article offers a review and comparison of existing methods of homomorphic encryption for machine learning tasks.

This paper addresses error correction codes approach in order to improve the performance of BOINC under uncertainty of users' behavior. Redundant Residue Number System (RRNS) moduli set of the special form provides correction of user unfairness, reliability, decreased redundancy and load of the computing network. Error correction code in RRNS is improved by using error syndrome. It decreases the amount of computations required for data decoding up to 20/7 times compared to the projection methods. The proposed modification of the error syndrome allow to omit the assumption that clients are honest and reliable. The proposed approach decreases the execution time of the client programs asymptotically by 4 times. On the other hand, encoding with RRNS places some restrictions on the range of performed operations.

Cloud computing has become a part of people's lives. However, there are many unresolved problems with security of this technology. According to the assessment of international experts in the field of security, there are risks in the appearance of cloud collusion in uncertain conditions. To mitigate this type of uncertainty, and minimize data redundancy of encryption together with harms caused by cloud collusion, modified threshold Asmuth-Bloom and weighted Mignotte secret sharing schemes are used. We show that if the villains do know the secret parts, and/or do not know the secret key, they cannot recuperate the secret. If the attackers do not know the required number of secret parts but know the secret key, the probability that they obtain the secret depends the size of the machine word in bits that is less than 1/2 ^(1-1) . We demonstrate that the proposed scheme ensures security under several types of attacks. We propose four approaches to select weights for secret sharing schemes to optimize the system behavior based on data access speed: pessimistic, balanced, and optimistic, and on speed per price ratio. We use the approximate method to improve the detection, localization and error correction accuracy under cloud parameters uncertainty.

In this paper, we propose a new approach to the construction of computationally secure secret sharing scheme based on the simultaneous use of the Residue Number System (RNS), as the symmetric encryption, and the perfect Asmuth- Bloom secret sharing scheme. We combine the useful features of Redundant RNS to design space efficient secret sharing scheme with enough level of security and to control of data integrity. This combination provides verifiability and achieves a high speed of data processing.