Machine Learning for Cybersecurity Cookbook

Private deep learning prediction

In many situations, company A might have a trained model that it wishes to offer as a service. At the same time, company A may be reluctant to share this model, so as to avoid having its intellectual property stolen. The simple solution to this problem is to have customers send their data to company A, and then receive from it predictions. However, this becomes a problem when the customer wishes to preserve the privacy of their data. To resolve this tricky situation, the company, as well as its customers, can utilize encrypted computation.

In this recipe, you will learn how to share an encrypted pre-trained deep learning model with a client and allow the client to predict using the encrypted model on their ...