The domestic startup Diostudio has recently unveiled a real-time lifestyle video anonymization technology based on generative AI technology. This solution transforms the faces of individuals in videos into animated or digital human styles in real-time, making personal identification impossible.

The technology is characterized by its ability to preserve the identity and individuality (facial expressions and behavioral characteristics) of the face while replacing it with a virtual face different from the actual person, thus rendering individuals in video data unrecognizable.

Diostudio explained, “By utilizing deep learning to convert to realistic virtual faces, the value as learning data is much higher than mosaic or blur processing.”

It has been pointed out that the typical blur or mosaic method significantly reduces the data value by damaging information necessary for AI learning during the process of obscuring human faces. In contrast, Diostudio's technology maintains useful information for learning, such as expressions, gaze, and behavioral context, while only removing personal identification elements, thus being evaluated as a new approach that satisfies both personal information protection and data utilization.

Experts believe that this new anonymization technology will create synergy with the government's 'K-Data Space' strategy and data utilization promotion policy. K-Data Space is a Korean data ecosystem strategy that aims for a decentralized data sharing system ensuring data sovereignty, interoperability, and trust-based exchange, similar to Europe's Gaia-X.

It is emerging as an alternative to overcome the limitations of centralized data collection by allowing individuals and companies to directly control their data and share or trade it safely when needed through a federated data platform.

The Digital Platform Government Committee and the Ministry of Science and ICT have also embarked on building an industrial data space to respond to the EU's Digital Product Passport (DPP). In such data spaces, personal information protection and security are crucial, and Diostudio's real-time face anonymization technology can be a key tool to safely share sensitive video data. This is evaluated as contributing to the realization of a digital platform government where public and private data flow seamlessly.

For the development of 'World models' and Physical AI models, considered national competitiveness in the AI era, vast high-quality learning data is essential. Applying this technology to video and image data, which could not be utilized due to personal information issues, can transform them into new data resources.

For AI models dealing with the physical world, such as autonomous driving, smart cities, and video security, pedestrian faces captured by street CCTV or in-vehicle cameras are needed. However, these data, difficult to utilize due to personal information regulations, can be safely used after anonymization processing. High-quality data, effectively anonymized while maintaining expressions or actions, is highly valuable as learning material for AI world models, i.e., super-large multimodal AI.

An industry insider emphasized, “The data space promoted by the government is ultimately intended to allow the private and public sectors to freely utilize data to secure data sovereignty and create innovative services,” adding that “privacy protection technologies like video anonymization must support this to facilitate the sharing of sensitive data and accelerate AI development.”

Diostudio is also seeking collaboration with the government and public institutions in line with this trend. Amid the surge in demand for face obscuration processing in videos due to strengthened regulations such as the Personal Information Protection Act, it is explained that they can support the utilization of video data in various fields such as education, healthcare, and transportation.

Diostudio CTO Park Sung-jin stated, “By using our technology, personal information can be anonymized while maintaining data quality, allowing institutions and companies to confidently engage in data sharing and integration.”

Popular News