Omsk University, in its new study, revealed that in the first quarter of 2025, approximately 801 million hacking attacks were launched on Russian companies, equating to more than a hundred attempts to access data within one second. It indicated that in our modern world, hackers target not only customers’ personal and financial data but also medical and biometric data.
To increase the level of protection for this information, scientists have developed a user voice authentication system based on a new neural network model. Pavel Lozhnikov, Deputy President of Omsk State Technical University for Research and Innovation, stated that the algorithm increased its ‘sensitivity’ to external interference thanks to new types of neural cells.
Lozhnikov confirmed that it was considered that a person could speak in a ‘natural’ voice, or be asleep, stressed, or tired. In one of the datasets used for training, speakers pronounced password phrases not only in a natural state but also in a modified state.
The university explained that the main problems this model can solve are the poor accuracy of voice recognition and maintaining the confidentiality of biometric models from intruders. In the future, Omsk Technical University scientists aim to solve the challenge of transferring the proposed model to other biometric features: handwriting, facial features, and others.
Experts also anticipate an increase in attacks on biometric systems using mannequins amidst the development of generative artificial intelligence and are conducting additional research to counter unauthorized access to this type of information.
