Off-Campus Hollins Users:
To access this document, please click here to log in to our proxy server with your campus network user name/password (the same one you use to log into the campus network and your e-mail).
Event Type
Research Presentation
Academic Department
Mathematics and Statistics
Location
Dana Science Building, 2nd floor
Start Date
14-4-2023 1:30 PM
End Date
14-4-2023 3:00 PM
Description
Under the direction of Weiqi Chu, University of California, Los Angeles (UCLA)
Advertisers in internet advertising access data about their ad campaigns across devices through secure, privacy-centric environments. To further improve the security protections for these environments, constraints against SQL queries involving multiple sets, or multi-sets, are in place. The purpose of this project is to develop efficient strategies and data structures to better detect multi-set overlaps and differences in order to better protect user privacy when evaluating multi-set queries. In this session, we will show our investigation into algorithms for detecting 3+ set differences. We developed two algorithms: the maximum-ID method and the reduced row echelon form (RREF) linear algebra-based method. The maximum-ID method uses element frequencies, or the number of times an element appears in a set, as a means to detect multi-set differences. The RREF method formulates queries and sets operations as vectors and matrices and converts the detection into a problem of matrix operations. For each method, we put forth theoretical analyses demonstrating privacy violations always caught coupled with experimental results highlighting additional potential use cases.
Detecting Small Multi-Set Differences Efficiently for Data Privacy
Dana Science Building, 2nd floor
Under the direction of Weiqi Chu, University of California, Los Angeles (UCLA)
Advertisers in internet advertising access data about their ad campaigns across devices through secure, privacy-centric environments. To further improve the security protections for these environments, constraints against SQL queries involving multiple sets, or multi-sets, are in place. The purpose of this project is to develop efficient strategies and data structures to better detect multi-set overlaps and differences in order to better protect user privacy when evaluating multi-set queries. In this session, we will show our investigation into algorithms for detecting 3+ set differences. We developed two algorithms: the maximum-ID method and the reduced row echelon form (RREF) linear algebra-based method. The maximum-ID method uses element frequencies, or the number of times an element appears in a set, as a means to detect multi-set differences. The RREF method formulates queries and sets operations as vectors and matrices and converts the detection into a problem of matrix operations. For each method, we put forth theoretical analyses demonstrating privacy violations always caught coupled with experimental results highlighting additional potential use cases.