Our lab focuses on Privacy-Enhancing Technologies (PETs) and AI Security, with the goal of enabling secure and trustworthy data analytics and AI systems. We develop practical techniques for secure data analysis in cloud and distributed environments using PETs such as homomorphic encryption, differential privacy, and federated learning. In addition, we conduct research on emerging AI security challenges, including LLM prompt security, retrieval-augmented generation (RAG) security, and AI agent security.
Privacy Enhancing Technologies (PETs)
- Homomorphic Encryption (HE)
- Differential Privacy (DP)
- Federated Learning (FL)
- Secure Biometric Recognition
AI Security
- Prompt Injection Defense
- Data Poisoning Defense
- Jailbraking Defense
- AI Red TeamingÂ