汪芷骐

PhD Student

College of IST, Penn State University

<aside> <img src="/icons/git_gray.svg" alt="/icons/git_gray.svg" width="40px" /> Github

</aside>

<aside> <img src="/icons/graduate_gray.svg" alt="/icons/graduate_gray.svg" width="40px" /> Google Scholar

</aside>

CV:

CV.pdf

<aside>

I am a first-year PhD student in the College of IST at Penn State University, advised by Professor Yuchen Yang.

Previously, I received my Bachelor’s Degree of Computer Science at RPI, where I was fortunate to be advised by Professor Lei Yu.

</aside>

💡Research Interests:

• ML Privacy: Machine Learning models should not reveal information beyond what's necessary to carry out the task. We want to minimize the risk of disclosing sensitive details to adversaries.
• Adversarial ML: Exploring methods to generate adversarial examples that intentionally manipulate machine learning systems to behave in ways that align with an adversary’s objectives, compromising the system’s intended functionality.

Topics I have investigated:

• Example difficulty in the image domain.
• Membership Inference Attacks to Traditional ML.
• Membership Inference Attacks to LLM.
• Adversarial prompts to LLM.
• LLM alignment faking.

Publications:

Membership Inference Attacks as Privacy Tools: Reliability, Disparity and Ensemble Z. Wang, C. Zhang, Y. Chen, N. Baracaldo, S. Kadhe, L. Yu To appear on ACM CCS 2025

GitHub - RPI-DSPlab/mia-disparity: This repository contains the source code for "Membership Inference Attacks as Privacy Tools: Reliability, Disparity and Ensemble", accepted by ACM CCS 2025.