Efficient Evaluation of Activation Functions over Encrypted Data

Efficient Evaluation of Activation Functions over Encrypted Data Patricia Thaine (University of Toronto) Presented at the 2nd Deep Learning and Security Workshop May 23, 2019 at the 2019 IEEE Symposium on Security & Privacy San Francisco, CA https://www.ieee-security.org/TC/SP2019/ https://www.ieee-security.org/TC/SPW2019/DLS/ ABSTRACT We describe a method for approximating any bounded activation function given encrypted input data. The utility of our method is exemplified by simulating it within two typical machine learning tasks: namely, a Variational Autoencoder that learns a latent representation of MNIST data, and an MNIST image classifier.