AF

Practical privacy-preserving machine learning methods are currently of critical importance in medical, financial and consumer applications, among others. The aim of this project is to develop practical private machine learning algorithms that can be easily implemented by practitioners in any field that holds sensitive data, while keeping robust privacy guarantees. The proposed research will extend the existing rigorous theoretical guarantees of differential privacy to reach the requirements of modern machine learning algorithms in concrete practical settings.

In the era of big data, across numerous application domains, large amounts of data are being collected, integrated, and analyzed to support rich application semantics and enable data-driven science discovery. While having the access to more data leads to more opportunities for analytics, mining, learning, and knowledge discovery, at the same time, it also increases the chance of security breach and privacy infringement.

Modern cryptography offers an impressive virtual buffet to a consumer who is wealthy in resources, with powerful tools like fully homomorphic encryption (which allows a provider to compute with encrypted values while keeping the client's data safe) and general purpose obfuscation (which allows one to hide the purpose of a given computation). But for more modestly minded users, who seek to perform less lofty tasks using more affordable computing resources or under more time-tested assumptions, the offerings are comparatively paltry.