Machine Learning and Practical Privacy

This course provides an introduction to research on privacy preserving machine learning systems. There are two main axis we include in terms of “practicality”. On one side of the practicality of a privacy-preserving system, we need to determine what it means to provide meaningful privacy to those whose data is used in such systems. On another side of the “practicality” included in this course, we need to ensure that technical guarantees are enforced and that they can be done with “reasonable” time and resources requirements. A practical privacy system must have sufficient utility. For instance, if a “privately trained” machine learning model does not achieve a certain level of success (e.g., at classification or generation) it is not able to serve its purpose.

The main assessment in this course is a project. It will include a project proposal, final project write up, and project presentation. The project is to be done in groups of 1-3 persons. The course will consist of lecture components as well as a seminar style component where students read a research paper in the field and present it to the class. More details can be found in the syllabus posted below.

Fall 2023 Offering

Lectures: Tuesday/Thursday 3:30-4:50pm

eClass: Access

Slides: 5Sept2023, 7Sept2023, 12Sept2023, 14Sept2023

Reference Syllabus: Draft Availabe Here