Assignment #3: Privacy-Preserving ML

Released on Friday 10/24/2025

Due Date: 23:59

Download [problems] [attachment]

Late Policy

You have free 8 late days.
You can use late days for assignments. A late day extends the deadline 24 hours.
Once you have used all 8 late days, the penalty is 10% for each additional late day.

Explore privacy-preserving machine learning techniques in this hands-on assignment. You will:

Privacy Attacks: Implement and evaluate membership inference attacks
Differential Privacy: Train models with differential privacy guarantees
Federated Learning: Implement a simple federated learning system
Privacy-Utility Tradeoff: Analyze the tradeoff between privacy and model performance
Evaluation: Compare different privacy-preserving approaches

Learning Objectives:

Understand privacy vulnerabilities in ML systems
Implement differential privacy mechanisms
Work with federated learning frameworks
Evaluate privacy-utility tradeoffs

Deliverables:

Implementation of privacy attacks and defenses
Experimental results and analysis
Report on privacy-utility tradeoffs

Resources: