[AI Revolution](32) DeepPingSource "Making All the World's Data Safe to Use"

AI Video Analysis Solution Without Privacy Infringement Concerns

With the rise of generative artificial intelligence (AI) such as ChatGPT, the demand for data across various industries is exploding. At the same time, concerns about personal information security are also increasing. DeepingSource develops de-identification technology and AI video analysis solutions that provide key information from video data without compromising personal privacy.

Founded in 2018, the startup DeepingSource offers AI video analysis solutions that do not raise privacy concerns. The technology anonymizes individuals in the footage so that they cannot be visually identified, similar to applying a mosaic effect, but AI can still extract essential information such as gender, age, movement paths, and gaze. For example, in a massive supermarket of about 4,000 to 5,000 pyeong (approximately 13,200 to 16,500 square meters), the system can predict the probability that a customer who bought apples and milk will proceed to the beer aisle and purchase beer, enabling real-time promotions like discounts accordingly. Kim Tae-hoon, CEO of DeepingSource, said, "The retail industry has long needed such solutions but hesitated due to privacy concerns. By adopting DeepingSource’s solution, customer data can be analyzed while protecting personal information."

Introduction to DeepPingSource De-identification Technology
Photo by DeepPingSource

CEO Kim is a co-founder and former Chief Technology Officer (CTO) of the image recognition startup Olaworks. After Olaworks was acquired by Intel in 2012, he was responsible for key projects such as developing AI processors at Intel. While working there, he recognized the business potential when the European Union (EU) implemented the General Data Protection Regulation (GDPR), which strengthened companies’ obligations to protect personal data. As obtaining images for AI training became extremely difficult, the demand for de-identified data surged.

When processing data for de-identification, faces are usually obscured with mosaic or blur effects to make them unrecognizable, or replaced with other faces using deepfake technology (AI-based image synthesis). However, in real-world applications, AI recognition accuracy may drop if the actual faces without mosaics are present. Synthesizing faces with deepfake technology is costly and difficult to implement. DeepingSource has developed proprietary de-identification technology that improves AI recognition accuracy while reducing costs. Additionally, images processed by DeepingSource cannot be restored to their original form, ensuring high security.

DeepingSource’s technology has attracted significant interest overseas, where privacy concerns are particularly sensitive. Since the enactment of the Data 3 Act in South Korea, domestic attention has also increased. The company currently holds over 80 patents in Korea, the United States, Europe, and Japan. The total investment amount has reached 23 billion KRW. CEO Kim stated, "We sold solutions to analyze the advertising effectiveness of exhibitions at COEX and digital signage. We also provide services to large supermarkets in the U.S. and are in discussions with multinational fashion companies. We have a competitive edge especially in large-scale spaces that are difficult for other companies to access."

While mainly adopted by retailers, the demand is rapidly expanding as AI applications spread across industries. The technology can be applied to logistics, industrial safety, banking, smart cities, cultural and commercial spaces, and more. For example, in autonomous driving, information such as pedestrians or license plates of other vehicles can be erased while enabling AI to distinguish people and vehicles from other objects. Recently, delivery robot and serving robot companies have also sought DeepingSource’s solutions.

CEO Kim aims to make all data in the world safely usable. Citizens can feel secure knowing that AI collects and analyzes data in daily life while protecting personal information, and companies can prevent unintended problems arising from using data containing personal information. He said, "I want to create a world where both data subjects and companies can use AI with peace of mind. Although we currently focus on video-based solutions, we plan to expand our business into other areas."