Home

CrCNN:A CNN Library Based On Homomorphic Encryption

The Crypto Convolutional Neural Network (CrCNN) library is a tool that allows to build CNNs able to make predictions homomorphically over encrypted images. The purpose of this library is to provide the building blocks to make experiments regarding the application of the HE to the field of Deep Learning, in particular on the CNNs. CrCNN satisfies these two main requirements:

• Requirement 1 Import the pre-trained model of a CNN and encode it.
• Requirement 2 Make predictions using the encoded CNN over encrypted data.