dwitch
durch TAE SEOK LEE
SaaS-type API service that identifies and anonymizes sensitive information within unstructured-data.
PIANO (Private Information ANOnymization) is a SaaS-type API service that identifies and anonymizes personal and sensitive information within unstructured data such as video, image, voice, and text.
It solves the problem of AI model performance degradation caused by existing anonymization techniques like simple blurring or masking. Dwitch utilizes Generative AI technology to preserve the usefulness of data while securely protecting personal information, achieving a balance between these two values.
Key Features
Video Anonymization: Automatically detects personal information (faces, license plates, etc.) frame by frame in a video and anonymizes it (using Gen AI, Blur, etc.).
Image Anonymization: Anonymizes personal information in image files and provides the results (image file, JSON).
Voice Anonymization: Detects personal information (names, contact numbers, etc.) through STT and applies Beep processing or modulates the voice.
Text Anonymization: Finds personal and sensitive information in text and masks it with asterisks (*) or other methods.
Technical Excellence & Differentiation
Next-Generation De-identification Tech based on Generative AI: Unlike traditional noise or blur methods, it uses Generative AI to maintain the original object's characteristics (gender, age group, etc.) while converting it into a virtual person/voice/text that cannot be identified as a specific individual.
Solves AI Training Performance Degradation: It maximizes the utility value of data by generating alternative information rather than simply deleting it. (When applying our Generative AI technology, the mAP (mean Average Precision) drop in AI models is minimized to 0.001).
Broad Scalability: Supports all types of unstructured data, including voice and text, not just images and videos.
Main Use Cases
Smart Factory: Worker privacy protection in factory CCTV footage.
Autonomous Driving: De-identification of personal information (faces, license plates) in video/image data for AI training.
Medical/Healthcare: Protection of personal information in medical data, such as surgical robot videos.
Public/Safety (Smart City): Securely utilizing intelligent CCTV footage as public data by anonymizing personal information.