Data Privacy (Fall 2026 draft) lab syllabus

Lab goals

The labs are where students move from vocabulary to actual technical reasoning. Each lab is designed to answer a concrete question:

Group policy

Format

Distribution

Lab notebooks and support files will be distributed separately through the course workflow, such as Canvas or a shared Drive folder. They are not published as raw files from this public website repository.

Lab 1: Privacy attacks on models

Theme: See the leak before you study the defense.

Main learning outcome: Students should be able to explain the difference between memorization, extraction, and inference attacks.


Lab 2: Re-identification and reconstruction

Theme: Privacy can fail even when names are removed.

Main learning outcome: Students should understand why de-identification alone is fragile and how statistical releases can still leak.


Lab 3: Private learning

Theme: Protecting training is not free.

Instructor note for the undergraduate version: the release should prioritize one clear core path over breadth. If needed, this lab can ship with a required core section plus one optional extension.

Main learning outcome: Students should be able to explain what the privacy budget buys, what it costs, and why implementation choices matter.


Lab 4: Secure multi-party computation (MPC)

Theme: Private computation under different trust assumptions.

Main learning outcome: Students should leave with the right mental model for when MPC is appropriate and where the performance bottlenecks come from.


Technical setup

Default platform

Software expectations

Instructional design guideline

For the undergraduate version, labs should reward interpretation and careful experimentation more than framework wrestling. If a toolchain becomes the main obstacle, the release should be simplified.