Diffprivlib - Privacy-Preserving Machine Learning with Scikit-Learn

Explore the concept of differential privacy and its application in machine learning through this 27-minute talk from EuroPython 2020. Learn how to integrate diffprivlib with scikit-learn and numpy to train accurate models with robust privacy guarantees. Discover the importance of data privacy in today's world and how to protect trained models from privacy vulnerabilities. Gain insights into mechanisms, models, tools, and budget accounting in privacy-preserving machine learning. Follow along with practical examples of running classifiers, building baselines, and creating histograms while maintaining data privacy. No prior knowledge of data privacy or differential privacy is required, but a basic understanding of scikit-learn is expected.

Intro
Data privacy
Diffprivlib approach
What is Diffprivlib
Mechanisms
Models
Tools
Accountant
Introduction to Diffprivlib
Running the classifier
Building a baseline
Import Diffprivlib models
Budget accountant
Preprocessing
Histogram
Results
Budget
Twodimensional histograms
Color maps
Queries
Additional Resources
Questions