Towards Practical Machine Learning with Differential Privacy and Variants

Опубликовано 9 января 2017, 22:07

Machine learning (ML) has become one of the most powerful classes of tools for artificial intelligence, personalized web services and data science problems across fields. However, the use of ML on sensitive data sets involving medical, financial and behavioral data are greatly limited due to privacy concern. In this talk, we consider the problem of statistical learning with privacy constraints. Under Vapnik's general learning setting and the formalism of differential privacy (DP), we establish simple conditions that characterizes the private learnability, which reveals a mixture of positive and negative insight. We then identify generic methods that reuse existing randomness to effectively solve private learning in practice; and discuss a weaker notion of privacy — on-avg KL-privacy — that allows for orders-of-magnitude more favorable privacy-utility tradeoff, while preserving key properties of differential privacy. Moreover, we show that On-Average KL-Privacy is **equivalent** to generalization for a large class of commonly-used tools in statistics and machine learning that sample from Gibbs distributions---a class of distributions that arises naturally from the maximum entropy principle. In addition, a byproduct of our analysis yields a lower bound for generalization error in terms of mutual information which reveals an interesting interplay with known upper bounds that use the same quantity. Finally, we describe an exciting future direction that allows us to bring together practitioners in census bureau, companies and academia to really advance the state-of-the-art for privacy protection.

See more on this video at microsoft.com/en-us/research/v...