The research of our lab covers various topics in trustworthy machine learning, such as robustness, privacy and interpretability, as well as their applications in computer vision (CV), natural language processing (NLP), cybersecurity and etc. Some of our current research projects are:
Fundations of Adversarial Machine Learning
- Understanding intrinsic robustness for inference-time/data-poisoning attacks
- Understanding robust overfiting of adversarial training
Attack/Defense against Adversarial Examples
- Semi-supervised methods for robust learning
- Robust certification methods
General Definitions of Model Robustness
- Cost-sensitive adversarial robustness
- Out-of-distrbution robustness
NLP Robustness
- Robustness Evaluation for Transformation-based Adversarial Defense