A research team at Chonnam National University led by Professor Yoo Seok-bong has created a protective technology called 'DeepProtect' aimed at preventing the misuse of deepfake technology, which is increasingly being exploited for unauthorized usage and digital crimes. This new technology offers a preemptive solution by securing facial images before they are uploaded online, making it challenging for deepfake applications to generate convincing results. Unlike conventional methods that primarily focus on post-detection of pre-existing deepfakes, DeepProtect emphasizes early intervention to maintain image integrity.

To bolster the effectiveness of the protection, DeepProtect employs a dual strategy of safeguarding the entire photo as well as specific features within it. The team has utilized techniques such as 'identity dilution,' which blends unique facial traits with those of similar faces to obscure overall identity information while maintaining vital characteristics. Additionally, they introduced a technique called 'attribute distortion' that allows the user to designate specific facial areas, such as the eyes, nose, and mouth, where minute watermarks can be inserted, rendering those features awkwardly represented in deepfake attempts.

Initial experimental results reveal that images treated with DeepProtect retain a high degree of naturalness in quantitative assessments and exhibit significant success rates against various advanced deepfake models. This research is co-authored by graduate students Baek Seung-hyuk and Lee Eun-gi and will be officially presented at the prestigious 'CVPR 2026 International Conference' in June, signaling a significant advancement in protective technologies against digital forgery and impersonation.