Research Topic #1
Architecture for an efficient privacy-preserving AI
중점 연구분야 #1
효율적인 프라이버시 보존 AI 아키텍처 연구
What should be protected ?
- Clients’ sensitive data: face image, health information…
- Model providers’ IP: trained sets of weight, bias and structure
How to efficiently protect ?
- Privacy-preserving using FHE and TEE
- Acceleration using FPGA architecture
Research Topic #2
Hardware-based TEE for confidential computing and architecture for AI-based intrusion detection
중점 연구분야 #2
기밀 연산을 위한 하드웨어 기반 TEE 연구와 지능형 침입 탐지 아키텍처 연구
Protected execution environment from outside
- Authentication and remote attestation based on Root-of-Trust
- Efficient computation running on FPGA parallel architecture
Host intrusion detection based on ML
- Training with normal states
- Trained model detects intrusion attacks in runtime