Анализ безопасности хранения биометрических данных с использованием FaceNet

Нано-био-технологии. Тепло- и электроэнергетика. Математическое моделирование: сборник статей III международной научно-практической конференции (Липецкий государственный технический университет, Липецк, Россия). 2025. С. 219–223.

Starodubov K., Гвасалия Г. В., Карасев П. И.

This paper discusses the concept of neural networks, convolutional neural networks, their architecture and their operation principle. The main attention is paid to testing the reliability of storing images of people as embeddings, which are considered to be unrecoverable in the original image. In the course of the research an experiment is carried out: the embedding of a face of a particular person is compared with a set of embeddings, which are stored in one file and among which there are both data of this person and data of other people. The accuracy of the embedding matching, the number of matches and the probability of identifying the person without accessing the original image are evaluated. An analysis of the results in the form of graphs and charts is derived to demonstrate the potential risk of information leakage and the vulnerability of this approach of storing human biometric data.

Keywords: машинное обучение, сверточные нейронные сети, Биометрические данные