MATH 388 - Statistical Data Privacy

Semester Offered: Fall or Spring
1 unit(s)

Statistical agencies and companies are under legal obligation to protect survey respondents’ and customers’ privacy when releasing respondent-level data and summary statistics to the public. Statistical methods could facilitate such release by introducing perturbation to the original confidential data. In this course, we explore two modern statistical data privacy approaches: 1) synthetic data, where statistical models are developed and respondent-level synthetic values are simulated from the estimated models for public release, and 2) differential privacy, a formal privacy definition that controls the amount of noise being injected mostly to summary statistics for public release. Students are expected to engage in independent learning and statistical programming, e.g., read accessible academic journal articles and implement learned methods with R or Python. Students develop an applied project that produces respondent-level data and/or summary statistics treated with statistical data privacy techniques, making it suitable for public release. Monika Hu.

Prerequisite(s): MATH 141 or MATH 240 (or equivalent); MATH 220, MATH 221, and MATH 241, or permission of the instructor.

Two 75-minute periods.

Course Format: CLS