AI Data and Security Research Center, Korea Electronics Technology Institute

Protecting AI models and data, and verifying the AI we deploy

We run edge robot AI inference confidentially in hardware-isolated space and protect the know-how inside models with irreversible learning and watermarking. We also verify supplied AI against its AI-BOM, a record of what it is built from, checking for vulnerabilities and mismatches.

Sensing and responding to on-site risks with video and sensors

We analyze video, sound and sensor data together to detect and respond to security and safety risks in buildings and industrial sites. We build security AI that runs on servers and lightweight edge devices, and an open video-security platform whose cameras protect themselves.

Understanding and linking data of different forms and sources

We build AI that analyzes video, sound, sensor data and text together, and verify the quality of multimodal retrieval-augmented generation. We also fuse heterogeneous data from scattered sources and share it robustly over unstable networks.

AI-generated digital humans, sound and 3D space for immersive content

We develop models that generate human appearance and motion, and technology that creates sound effects and ambience to match a scene. We also reconstruct low-visibility spaces in 3D and personalize hearing and touch across different XR devices.