SUSU Scientists Ensure Data Protection in Cloud StorageMay 12, 2021
Within the framework of the Year of Science and Technology in Russia, each month has its own specific topic. May is dedicated to security issues, with a special focus on the digital sphere.
Due to the annually growing demand for cloud storage, addressing the issue of data protection is of particular importance. In modern storages, data is stored in encrypted form. However, to process and analyze information, it is necessary to decrypt it, and this significantly increases the risk of leakage.
SUSU specialists from the Research Laboratory for Problem-Oriented Cloud Environments (Higher School of Electronics and Computer Science) conducted fundamental research in the study of data protection issues in cloud storage. The research is based on the ability to process data without the need to decrypt it. This process will allow for secure processing of protected data on open cloud resources by neural networks.
The key aspect for the university scientists was the implementation of their own methods and approaches for solving the machine learning problem when working with encrypted data. Work in this area is carried out by the head of the laboratory, Professor Andrei Chernykh, and postdoc Jorge Mario Cortez-Mendoza.
“One of the most frequent examples being solved by data processing virtual centres are data processing tasks by methods of computer-assisted instructions, in particular by neural nets. Standard methods of data processing by neural nets require decoding. In our initial research we are analyzing existing approaches that allow to ensure data processing by neural net means without decoding”, — says Gleb Radchenko, Director of School of Electronic Engineering and Computer Science, the South Ural State University.
Besides, the scientists of the South Ural State University together with their colleagues from such universities as CICESE, (Mexico), Ivannikov Institute for System Programming of the RAS (Moscow, Russia), North-Caucasus Federal University (Stavropol, Russia), MIPT (Moscow, Russia)) for the first time conducted work on comparison research among existing neural nets that are using the main method of homomorphic coding which does not require data decoding while dealing with it. The published work describes fundamental concepts of the data processing systems without their decoding required, discovered modern working tools, described existing open problems and their solutions connected with the homomorphic coding and computer-assisted instructions. This research will help not only to understand the essence of the set tasks with the help of new implemented technologies and algorithms but also to improve theoretical basis.