Recent history has shown that the benefits brought forth by today’s data driven culture come at a cost, for example the Yahoo and Equifax data leakages. Such breaches have a devastating impact on individuals and industry, and lead the community to seek for privacy preserving solutions. We discuss one potential approach, which enables machine learning over encrypted data, and thus providing resiliency against information leakage.

Several machine learning algorithms have already been adapted to work over encrypted data, including a solution developed at Intuit for training and evaluating decision trees with strong privacy guarantees. We discuss open problems that need to be addressed by the industry for successful adaption of this approach.