UDC 346.24/.26 (476)
Authors
Bekoeva A. A.
Bachelor of the third year of study in Economics,
North Ossetian State University named after K.L. Khetagurova, Vladikavkaz, Russia
Abstract
Currently, South Korea’s privacy laws are designed to respond to
the public demand that the right to self-determination of personal information
be properly protected, while allowing the use of personal information for the
purposes of training artificial intelligence. De-identification measures, such as
pseudonymization and anonymization, have been legislated as a solution to this
problem, but criticism that they are not sufficient to protect personal information,
as they do not sufficiently reduce the risk of re-identification, and the excessive
use of security technologies significantly reduces the usefulness of data,
is still significantly increases. As an alternative, a new technical approach that
goes beyond de-identification measures has recently attracted attention. This
research is a technical approach that allows artificial intelligence to learn and
reason, while strengthening the protection of personal information in the field of
artificial intelligence. Simultaneously with the application of this approach, the
significance and limitations of each technology are considered.
1) Data transformation approach (differential privacy);
2) distributed data minimization approach (federated learning);
3) approaches to data encryption (homomorphic encryption);
4) approach to the creation of virtual data (synthetic data), this study
suggests that privacy laws should evolve in a direction that embraces these new
technologies and enhances both the level of personal information protection and
the usefulness of data.
Keywords
privacy protection technology, differential privacy, federated learning, homomorphic encryption, synthetic data.