To obtain access to full text of journal and articles you must register!
- Article name
- Technology of "blind" processing of attracted data in machine learning systems
- Authors
- Konyavsky V. A., , konyavskiy@gospochta.ru, Moscow Institute of Physics and Technology (State University); Federal State Budgetary Research Institution "The State Scientific and Technological Center "Science", Dolgoprudny, Moscow Region, Russia; Moscow, Russia
Ross G. V., , ross-49@mail.ru, Financial University under the Government of the Russian Federation, Moscow, Russia
Konyavskya-Schastnaya S. V., , cd@okbsapr.ru, Moscow Institute of Physics and Technology (National Research University); SC "OKB SAPR", Moscow Region, Dolgoprudny, Russia; Moscow, Russia
Raigorodskii A. M., , mraigor@yandex.ru, Moscow Institute of Physics and Technology (National Research University); Lomonosov Moscow State University; Caucasus Mathematical Center, Adyghe State University; Banzarov Buryat State University, Moscow Region, Dolgoprudny, Russia; Moscow, Russia; Republic of Adygea, Maykop, Russia; Ulan-Ude, Buryat Republic, Russia
Trenin S. A., , s.trenin@gmail.com, Moscow Institute of Physics and Technology (National Research University), Moscow Region, Dolgoprudny, Russia
Leonidov A. V., , leonidovav@lebedev.ru, Moscow Institute of Physics and Technology (National Research University); P. N. Lebedev Physical Institute of the Russian Academy of Sciences, Moscow Region, Dolgoprudny, Russia; Moscow, Russia
Vasilyeva E. E., , serebryannikovaee@lebedev.ru, Moscow Institute of Physics and Technology (National Research University); P. N. Lebedev Physical Institute of the Russian Academy of Sciences, Moscow Region, Dolgoprudny, Russia; Moscow, Russia
Vasilyev S. B., , svasilev@hse.ru, Moscow Institute of Physics and Technology (National Research University); P. N. Lebedev Physical Institute of the Russian Academy of Sciences; National Research University "Higher School of Economics", Moscow Region, Dolgoprudny, Russia; Moscow, Russia
Konovalyhin M. Yu., , konovalihin@vtb.ru, VTB Bank (PJSC), St. Petersburg, Russia
- Keywords
- machine learning / security / "blind" data processing / confidential machine learning / IT Pipeline / data combining / data that is to be protected / unretrievable data / direct leaks / indirect leaks / differential attacks / confidential computing / national antifraud operator
- Year
- 2024 Issue 2 Pages 17 - 32
- Code EDN
- ICEKIG
- Code DOI
- 10.52190/2073-2600_2024_2_17
- Abstract
- The article is devoted to resolving the contradiction that has arisen to date between the need to process combined data when building predictive models and the lack of technical solutions to ensure a sufficient level of information security. This created a contradiction between the need to process combined data and the lack of technical solutions to ensure a sufficient level of information security. This article is devoted to resolving this contradiction. The concepts of blind data processing, unretrievable data and models, and attracted data are introduced. Examples of indirect leaks are given, the occurrence of which is usually associated with differential attacks. To increase the level of data security, it is proposed to limit requests to the machine learning system during blind data processing only to proven command sequences that do not lead to leaks of data that is to be protected - IT pipelines. The formal model of blind data processing, the reference architecture of an automated blind data processing system, the structure and measures for fixing and ensuring the integrity and legality of the use of IT pipelines, features of their formation and application are proposed. Based on the results obtained, the Crypto-Enclave software and hardware complex was created. The results can also be used in creating a national antifraud system, in medicine, and industry.
- Text
- BUY for read the full text of article
- Buy
- 500.00 rub
