Hancomwith, a subsidiary of Hancom Group, announced on the 8th that it has signed an industry-academic cooperation agreement with the AI Security Research Center (AISRC) of Soongsil University for the joint research and development of 'Deep Voice detection and voice authentication technology.' Through this collaboration, Hancomwith plans to introduce new products utilizing Deep Voice detection and voice authentication technology either independently or in combination.
Voice authentication technology extracts and stores phonetic information based on the user's free speech and verifies identity by comparing it with the registered information during a currency. This provides a faster and more convenient authentication process than the traditional question-answer method. With advancements in AI technology, voice has become a key interface, and the use of voice-based authentication and identification technologies is expected to expand in various fields such as finance and public services.
The Deep Voice detection system developed by the AI Security Research Center of Soongsil University receives voice files as input and utilizes advanced deep learning technology to distinguish between human voices and synthetic voices. It supports various environments such as web, mobile, and PC, achieving over 97% detection performance on the fake voice dataset (ASVspoof 2021 DF) and 94% detection performance on data collected in real-world environments (In-The-Wild). This performance exceeds that of the currently available state-of-the-art (SOTA) models.
Deep Voice is an AI-based voice synthesis technology that can accurately replicate a specific person's voice with just a small number of voice samples. Recently, the number of voice phishing crimes using the voices of corporate chief executive officers (CEOs), celebrities, and family members has increased, expanding the scale of damage. Accordingly, Deep Voice detection and voice authentication technology are expected to serve as effective countermeasures against crimes like voice phishing, contributing to reducing social and economic damage.
Song Sang-yeop, the representative of Hancomwith, said, 'As Deep Voice technology becomes more sophisticated with the advancement of AI, the importance of technology that can distinguish between real and fake is growing.' He added, 'We will expand our competitiveness in the security market through Deep Voice detection and voice authentication technology, which has high security and reliability.'