Yingtai Xiao
Research Scientist · Privacy Innovation Lab @ TikTok
> researching differential privacy
01About Me
My name is Yingtai Xiao. I am a Research Scientist in the Privacy Innovation Lab at TikTok. Previously I earned my Ph.D. degree in Computer Science and Engineering at The Pennsylvania State University, advised by Prof. Daniel Kifer and Prof. Danfeng Zhang.
I received my B.S. degree in Mathematics from the University of Science and Technology of China, where I was supervised by Prof. Juyong Zhang. I also visited Nanyang Technological University as a research assistant, under the supervision of Dr. Yong Liu.
I’m interested in differential privacy and its applications in database query optimization, advertising measurement and machine learning. For a list of my publications, see Google Scholar.
02Research Interests
Differential Privacy
Mechanism design with provable privacy guarantees and optimal utility.
Database Query Optimization
Answering workloads of linear queries accurately under a privacy budget.
Advertising Measurement
Production-scale ad measurement under user-level differential privacy.
Private Machine Learning
Training and synthesizing programs that respect privacy by construction.
03Activities
- Jun 2024
Joined the Privacy Innovation Lab at TikTok as a Research Scientist.
- May 2024
Graduated with a Ph.D. in Computer Science and Engineering from Penn State.
- Summer 2023
Internship at TikTok — built an online advertising measurement system under user-level differential privacy.
- Summer 2022
Internship at JD.com — built a three-party secure multiparty computation framework for deep learning.
- Fall 2021
Head teaching assistant of CMPSC 461 Programming Language Concepts.
04Selected Publications all papers
- NeurIPS An Optimal and Scalable Matrix Mechanism for Noisy Marginals under Convex Loss FunctionsConference on Neural Information Processing Systems (NeurIPS 2023) Spotlight
- VLDB Answering Private Linear Queries Adaptively using the Common MechanismProc. of the 49th International Conference on Very Large Data Bases (VLDB 2023)
- VLDB Optimizing Fitness-For-Use of Differentially Private Linear QueriesProc. of the 47th International Conference on Very Large Data Bases (VLDB 2021)
- CCS DPGen: Automated Program Synthesis for Differential PrivacyProc. of the ACM Conference on Computer and Communications Security (CCS 2021)