Thursday, September 27, 2018

On September 7, 2018, Alberto Ibarrondo (EURECOM) presented the paper FHE-compatible Batch Normalization for Privacy Preserving Deep Learning at the 13th Data Privacy Management International Workshop (DPM 2018) in Barcelona, Spain, during the session Privacy and Cryptography.

The paper proposes a batch normalization layer compatible with fully homomorphic encryption for deep neural networks that improves performance of classification with no loss on accuracy.

This work is supported by PAPAYA.