14 November 15:09
Russian businesses have three scenarios for working with big data and AI: process information in trusted environments, unite for joint computing, or invest in homomorphic encryption.
A detailed analysis of each strategy was conducted at the Digital Solutions forum by Sergei Karpovich, deputy head of the T1 AI department (part of the T1 IT holding).
Speaking at the session “Innovations in Privacy: New Tools for Working with Big Data,” Sergei Karpovich noted the rapid growth of the confidential computing market. If just a few years ago its volume was estimated at several billion rubles, then by the 2030s it is predicted to reach 100 billion. The growth is primarily due to increased requirements for data security – this topic has become a priority not only for Russian, but also for global business. The dynamics of the global market is plus 25% annually.
Companies are increasing investments in information security technologies, driven by tightening regulations, and are looking for new approaches to ensuring privacy. Enterprises no longer want to provide their own data to create analytical cases and models – on the contrary, market participants are increasingly “sovereigning” the processing of customer information.
However, this comes with a number of barriers, including high investment costs. According to Sergei Karpovich, server complexes will cost 15–20 million rubles. In addition, operating modern IT solutions requires a highly qualified team. All this makes confidential computing technologies inaccessible, especially for SMEs.
Sergey Karpovich sees three possible scenarios in which companies of all sizes and formats will be able to work effectively with big data.
First, moving computing to trusted cloud environments where data can be stored and processed without operators having direct access to the content.
Secondly, the development of secure collaborative computing technologies (Secure Multi-Party Computation, SMPC). In this case, several organizations with similar data can jointly form analytical models without disclosing the source data to each other. In the banking sector, similar technologies are already used to build collective risk models and jointly analyze information when exchanging encryption keys. Distributed storage of key fragments among different owners allows for a high level of protection: only sharing the entire set of keys makes it possible to decrypt data if necessary.
The third promising direction is homomorphic encryption, which makes it possible to perform calculations inside encrypted data arrays. Although such technologies are still far from mature due to the complexity of the encryption procedure itself and subsequent decryption, there is already experience in their use – for example, in solutions for electronic voting and when processing information from new devices.
“In the next two years, confidential computing technology will become one of the key technologies shaping a new technological order. An analysis of the advantages and limitations of various technologies shows that they complement each other and are suitable for solving specific problems. The key task of market participants is to adequately assess the applicability of each approach and choose the most effective solution for the best protection, analysis and use of data,” Sergey Karpovich is confident.
