Research

Research Focus

Our research covers three levels of trustworthy AI — foundational theory, algorithm design, and real-world system deployment.

120+

Published Papers

50+

CCF-A / Top-tier

4

Best Paper Awards

1660+

Top Citation (MAD-GAN)

🛡

LLM Security & Safety Alignment

  • LLM safety evaluation & red teaming
  • Jailbreak attack detection & defense
  • Backdoor attacks & adversarial robustness
  • Quantization-aware safety patching
  • Prompt watermarking & copyright protection

PromptCARE (S&P'24), PoisonedEye (ICML'25), ACQ (MM'23)

🔐

Privacy Protection & Federated Learning

  • Differential privacy theory & applications
  • Personalized federated learning
  • Secure multi-party computation (MPC)
  • Machine unlearning
  • Privacy-preserving data sharing

CCS'24 Distinguished Paper, VLDB'22, NeurIPS'23, ICCV'23

📈

Time Series Intelligence & Industrial AI

  • Multivariate time series anomaly detection
  • Predictive maintenance & fault diagnosis
  • Time series foundation models
  • Time series data quality assessment
  • Industrial IoT data governance

MAD-GAN (1660+ citations), MAD-SGCN (ICDE'22), ICLR'26

💎

Data Markets & Valuation

  • Data valuation mechanisms
  • Fair data pricing
  • Data rights & right to be forgotten
  • End-to-end data marketplace infrastructure
  • Data compliance analysis

Dealer (VLDB'21), Equitable Data Valuation (VLDB'23)

🧪

Data Quality & Governance

  • Data quality assessment & repair
  • Missing value imputation & augmentation
  • DP synthetic data generation
  • Cross-domain transfer learning
  • Representation learning & heterogeneous fusion

IGAMT (AAAI'24), CB-GAN (DASFAA'23)

🤖

Trustworthy Agents & RAG Systems

  • RAG system reliability analysis
  • Knowledge poisoning attack & defense
  • Autonomous agent safety evaluation
  • Multimodal safety evaluation
  • Trustworthy AI for databases

PoisonedEye (ICML'25), SQL Injection via Backdoor (SIGMOD'26)