Scientists from the Umsk State Technical University in Russia have developed an innovative model of a nervous network aimed at protecting personal and confidential data from leaks, through a authentication system based on audio identification, taking into account the natural changes in the tone of the sound and its condition according to the psychological situation of the user.
The university stated, in a statement, that the new system comes in response to the growing threats of cybersecurity, as the first quarter of 2025 witnessed about 801 million piracy attacks on Russian companies, at a rate of more than a hundred illegal access to data per second.
The developer model allows accurately verification of the user’s identity through his voice, while enhancing protection against attempts to cloning voice passwords. It is characterized by a higher ability to resist external overlap thanks to new types of neurons and sports bonds, and the accuracy of identification is a better percentage of similar models (the probability of error 2.1% compared to 2.7%), in addition to the fact that the password generated reaches 1024 bits compared to 1660 bits in the traditional models.
Pavel Luznikov, Vice President of the University for Research and Innovation, stressed that the system was subject to experiments that included sound recognition in different circumstances, such as fatigue, tension or sleep, to ensure the accuracy of performance in all conditions.
The statement pointed out that the new technology can extend in the future to include other biometric patterns, such as handwriting and facial features, in light of expectations of increasing attempts to penetrate biological measurements using obstetric intelligence technologies, which requires the development of advanced solutions to counter these threats.
