Abstract

Differential privacy approaches employ a curator to control data sharing with analysts without compromising individual privacy. The curator’s role is to guard the data and determine what is appropriate for release using the parameter epsilon to adjust the accuracy of the released data. A low epsilon value provides more privacy, while a higher epsilon value is associated with higher accuracy. Counting queries, which ”count” the number of items in a dataset that meet specific conditions, impose additional restrictions on privacy protection. In particular, if the resulting counts are low, the data released is more specific and can lead to privacy loss. This work addresses privacy challenges in single-attribute counting-range queries by proposing a Workload Partitioning Mechanism (WPM) which generates estimated answers based on query sensitivity. The mechanism is then extended to handle multiple-attribute range queries by preventing interrelated attributes from revealing private information about individuals. Further, the mechanism is paired with access control to improve system privacy and security, thus illustrating its practicality. The work also extends the WPM to reduce the error to be polylogarithmic in the sensitivity degree of the issued queries. This thesis describes the research questions addressed by WPM to date, and discusses future plans to expand the current research tasks toward developing a more efficient mechanism for range queries.

Publication Date

6-2020

Document Type

Dissertation

Student Type

Graduate

Degree Name

Computing and Information Sciences (Ph.D.)

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Rajendra K. Raj

Advisor/Committee Member

Carol J. Romanowski

Advisor/Committee Member

Sumita Mishra

Campus

RIT – Main Campus

Share

COinS