CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
PROTECTION METHODS IN MODERN VOICE AUTHENTICATION SYSTEMS
 |
Read | Evsyukov Michael V., Putyato Michael M., Makaryan Alexander S., Nemchinova Valeriya O. PROTECTION METHODS IN MODERN VOICE AUTHENTICATION SYSTEMS // Caspian journal : management and high technologies. — 2022. — №3. — pp. 84-92. |
Evsyukov Michael V. - Kuban State Technological University
Putyato Michael M. - Kuban State Technological University
Makaryan Alexander S. - Kuban State Technological University
Nemchinova Valeriya O. - Kuban State Technological University
The problem of improving the security of voice authentication systems is becoming increasingly important due to the rapid development and widespread use of voice interfaces. Although voice recognition algorithms are wellstudied and demonstrate high reliability when tested by live people, modern voice authentication systems are subject to a number of vulnerabilities. First of all, this is due to the access to affordable and high-quality devices for recording and playing sound. This fact provides attackers with powerful tools to implement attacks on voice authentication systems. As a rule, the attacker's purpose is to authenticate in the system under the guise of another person. Actions aimed at achieving this goal are called spoofing. This article describes the main voice characteristics used in the implementation of voice authentication systems, provides an up-to-date classification of voice recognition algorithms, describes existing metrics for evaluating the effectiveness of voice authentication systems, and outlines existing approaches to classifying spoofing methods. In addition, the classification of countermeasures against spoofing has been improved and promising directions for future research in the field of voice authentication have been identified.
Key words: biometrics, artificial intelligence, machine learning, information security, authentication, spoofing, liveness detection