Hancomwith Launches AI-Driven Security to Combat Voice Deepfakes

Published Date : 10/22/2025 

Hancomwith, a subsidiary of Hancom Group, has launched 'Hancom AISRC Speaky,' an AI-driven security solution that combines voice verification with deepfake voice detection technology. This innovative tool is part of a broader effort to combat audio-based identity fraud. 

Hancomwith, a subsidiary of Hancom Group, has unveiled 'Hancom AISRC Speaky,' an advanced AI-powered security solution that integrates voice verification with deepfake voice detection technology. This release marks the first commercial outcome of a collaborative R&D initiative between Hancomwith and the AI Security Research Center (AISRC) at Soongsil University in Seoul.

The partnership, established in April, is focused on developing cutting-edge technologies for voice authentication and the detection of synthetic audio. AISRC has been awarded six billion South Korean Won ($4.2 million) by the South Korean government for an eight-year research project aimed at enhancing system security solutions and defending against cyber threats.

'Speaky' is an innovative authentication tool designed to replace traditional methods such as passwords and public key certificates. By analyzing nuanced vocal characteristics, including individual intonation and regional dialects, the system can effectively identify and block spoofing attempts that rely on AI-generated voices.

According to Hancomwith, the solution leverages a deep learning-based voice recognition model trained on extensive Korean-language datasets. It also incorporates a proprietary 'triple training strategy' aimed at minimizing data bias and enhancing the model's adaptability across various environments. This approach has reportedly resulted in a robust and reliable detection performance.

'We plan to continue introducing innovative next-generation AI security technologies, including passive authentication,' said Song Sang-yeop, CEO of Hancomwith. The launch of Speaky is part of a growing industry effort to address the rising threat of audio-based identity fraud, particularly as generative AI tools make it easier to fabricate convincing synthetic voices.

Hancomwith's new solution is a significant step forward in the field of voice biometrics and deepfake detection, offering a more secure and reliable method for verifying identities and protecting against fraudulent activities. 

Frequently Asked Questions (FAQS): 

Q: What is Hancomwith's 'Hancom AISRC Speaky'?

A: Hancomwith's 'Hancom AISRC Speaky' is an AI-powered security solution that combines voice verification with deepfake voice detection technology. It is designed to replace traditional authentication methods and enhance security against AI-generated voice spoofing.

Q: What is the main purpose of the Speaky tool?

A: The main purpose of the Speaky tool is to provide a more secure and reliable method for voice authentication and to detect and block attempts to use AI-generated voices for identity fraud.

Q: How does Speaky work to detect deepfake voices?

A: Speaky works by analyzing nuanced vocal characteristics such as individual intonation and regional dialects. It uses a deep learning-based voice recognition model trained on extensive datasets to identify and block spoofing attempts.

Q: What is the 'triple training strategy' mentioned in the article?

A: The 'triple training strategy' is a proprietary approach used by Hancomwith to minimize data bias and enhance the adaptability of the voice recognition model across diverse environments, resulting in more robust and reliable detection performance.

Q: Why is the Speaky solution important in the context of rising audio-based identity fraud?

A: The Speaky solution is important because it addresses the growing threat of audio-based identity fraud, which has become more prevalent with the advancement of generative AI tools that can easily fabricate convincing synthetic voices. By providing a more secure authentication method, Speaky helps protect against such fraudulent activities. 

More Related Topics :