Practical Data Analytics under Differential Privacy

Speaker: Dr. Wei DONG
         PostDoc Fellow, Carnegie Mellon University
         and
         CSE Best PhD Dissertation Award (2022-23) Recipient
         Hong Kong University of Science and Technology

Title:  "Practical Data Analytics under Differential Privacy"

Host:   Prof. Kai Chen

Date:   Monday, 6 May 2024

Time:   4:00pm - 5:00pm

Venue:  Lecture Theater F (Leung Yat Sing Lecture Theater)
        near lift 25/26, HKUST

Abstract:

In the big data era, organizations continuously collect vast amounts of 
sensitive information, and a key challenge is to get meaningful analytical 
results without compromising privacy. As a gold standard for private data 
analysis, differential privacy (DP) has garnered significant attention 
from both academia and industry. Informally speaking, DP requires that 
query results are indistinguishable regardless of whether any particular 
individual's data is included or not in the database thus we cannot infer 
any individual's information through the query result. Noise injection is 
inherently necessary for this goal. Although DP is widely researched in 
all kinds of data science areas, many DP mechanisms do not provide a 
practical utility (error level) in real-world applications. The primary 
challenge is that traditional DP methods often set a universal limit on 
individual contributions to queries and add noise in proportion to this 
limit. This results in a consistent error level on every instance, which 
can be very high in practice. To address this, I advocate for a novel 
concept: instance-optimal error, which minimizes error for each specific 
instance rather than universally. This "paradigm shift" in DP design 
enhances practical utility significantly. My focus is on two areas: 
developing a practical DP SQL query engine and designing practical DP 
solutions for artificial intelligence.


*******************
Biography:

Dr. Wei Dong is a Postdoctoral Fellow in the Department of Computer 
Science at Carnegie Mellon University. He received the Ph.D. degree from 
the Hong Kong University of Science and Technology (HKUST). His general 
areas of interest include data security and privacy, database theory and 
algorithms, and machine learning. His research has been recognized by the 
academic community and appeared in top-tier conferences, such as SIGMOD, 
S&P, PODS, CCS, VLDB, NeuIPS, and KDD. He received the Best Paper Award in 
SIGMOD 2022 (1/514) and SIGMOD 2023 Research Highlight Award. He also 
received the 2023 Ph.D. Research Excellence Award from the School of 
Engineering at HKUST.