This is a practical tutorial based on Adam Smith Privacy in Statistics and Machine Learning Course.
I will try to explain the math using examples and python code. This will help beginners to get better understanding! The original course includes video lectures and slides so I suggest you start there and then practice here. Each topic will have its own colab notebook. All notebooks will be available publicly in github. You can use the [github discussion] section to ask questions and hopfully getting answers.
Notes:
- I will use photos and resources from these lectures and other resources, all source links will be available.
- This is still work on progress. I will try to add more contents whenever I have more free time
- Contribution is welcome!
- If you have questions, comments, corrections, or feedback, please post in the dicussion forum.
- Introduction
- Reconstruction Attacks
- Differential Privacy Fundamentals
- Exponential Mechanism and Report Noisy Max
- The Binary Tree Mechanism
- Approximate DP
- Advanced Composition
- Private Empirical Risk Minimization
- Private Gradient Descent
- Factorization Mechanisms
- The Projection Mechanism
- Online Learning and Multiplicative Weights
- Synthetic Data Generation and Online Learning
- Two-Player Zero-Sum Games
- Synthetic Data in Practice
- The Limits of Privacy
- Private Statistical Inference
- Privacy and Adaptive Data Analysis
- Differential Privacy and the Census