Hello! I am a PhD student in the School of Computing at Macquarie University, jointly supervised by Prof. Mark Dras and Dr. Qiongkai Xu. My primary research interest is in tackling privacy and security challenges in ML/NLP models and their applications.
I approach these challenges from two directions:
- Data privacy: auditing potential leakage throughout the lifecycle of machine learning models (e.g., reconstruction attacks) and evaluating or improving defenses such as differential privacy.
- Security: investigating mechanisms that make models vulnerable to adversarial manipulation, particularly backdoor attacks, and developing interpretable defenses to enhance robustness and trustworthiness.
Beyond these areas, I am keen to explore advanced and emerging topics in privacy and security, and I welcome opportunities for collaboration on related challenges.
Latest News
- [Oct 2025] Our paper “Securing Off-the-Shelf LLMs against Backdoor Attacks” has been accepted for presentation at ALTA 2025. Looking forward to seeing you in Sydney in November!
- [Sep 2025] I am admitted to attend Machine Learning Summer School (MLSS) Melbourne 2026. Looking forward to seeing you in Melbourne in February!
- [Sep 2025] Our competition proposal on Anti-BAD Challenge has been accepted by IEEE SaTML 2026.
- [Aug 2025] Our paper Cut the Deadwood Out: Backdoor Purification via Guided Module Substitution has been accepted to Findings of EMNLP 2025.