AIRI: Synthetic Voice Definition System

2024: Product Announcement

In September 2024, specialists from the Institute of Artificial Intelligence AIRI and the Moscow Technical University of Communications and Informatics (MTUSI) presented the first AI system in Russia capable of recognizing voice deepfakes by phone. This development is aimed at combating telephone fraud, and can also improve the security of systems using voice authentication.

According to TASS, the system is based on the use of artificial intelligence, which analyzes audio sequences to identify signs of synthetic voice. The principle of operation is that modern AI is able to take into account the context of the conversation and the features of voice data, which allows it to distinguish between genuine and generated voices with high accuracy. This statement was made by the head of the scientific group "Trusted and Secure Intelligent Systems" AIRI Oleg Rogov. He also noted that the new system is significantly superior in performance to pre-existing foreign counterparts, including the development of scientists from South Korea and France.

𝓲

Freepik

Fraudsters are increasingly using text-to-speech and voice-processing technologies to create deepfakes that are difficult to distinguish from real voices. Russian scientists have proposed a solution that can effectively cope with such threats. The developed system not only detects fakes, but also adapts to new types of attacks, which makes it a universal tool for ensuring security in various areas, including telephone communications and voice authentication systems.

According to Grach Mkrtchyan, head of the Intellectual Solutions Research Department of MTUSI, the technology can be integrated into existing security systems both as separate software and as a module for existing systems. The introduction of such a tool significantly reduces the risk of telephone fraud, as well as improves the protection of personal data in systems using voice technologies.^[1]

Notes