Privacy-preserving Analytics and Machine Learning with Differential Privacy

QuantUniversity Spring School 2021 Session by Andreas Kopp and Lucas Rosenblaat Pioneered by Microsoft Research and associates, differential privacy is the emerging gold standard for protecting data in applications like preparing and publishing statistical analyses. Differential privacy provides a mathematically measurable privacy guarantee to individuals by adding a carefully tuned amount of statistical noise to sensitive data. It promises significantly higher privacy protection levels than commonly used disclosure limitation practices like data anonymization. What differential privacy is and how it works Microsoft's and Harvard's OpenDP initiative and the SmartNoise system Using SmartNoise to protect sensitive data against privacy attacks How to create differentially private synthetic data using the new SmartNoise synthesizers Performing analytics, machine learning including deep learning on sensitive data using differential privacy The trade-off between privacy guarantee and accuracy of analytical results How to engage in our Early Adopter program