English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

The adoption of machine learning solutions is rapidly increasing across allparts of society. As the models grow larger, both training and inference ofmachine learning models is increasingly outsourced, e.g. to cloud serviceproviders. This means that potentially sensitive data is processed on untrustedplatforms, which bears inherent data security and privacy risks. In this work,we investigate how to protect distributed machine learning systems, focusing ondeep convolutional neural networks. The most common and best-performing mixedMPC approaches are based on HE, secret sharing, and garbled circuits. Theycommonly suffer from large performance overheads, big accuracy losses, andcommunication overheads that grow linearly in the depth of the neural network.To improve on these problems, we present Dash, a fast and distributed privateconvolutional neural network inference scheme secure against maliciousattackers. Building on arithmetic garbling gadgets [BMR16] and fancy-garbling[BCM+19], Dash is based purely on arithmetic garbled circuits. We introduceLabelTensors that allow us to leverage the massive parallelity of modern GPUs.Combined with state-of-the-art garbling optimizations, Dash outperformsprevious garbling approaches up to a factor of about 100. Furthermore, weintroduce an efficient scaling operation over the residues of the Chineseremainder theorem representation to arithmetic garbled circuits, which allowsus to garble larger networks and achieve much higher accuracy than previousapproaches. Finally, Dash requires only a single communication round perinference step, regardless of the depth of the neural network, and a very smallconstant online communication volume.

Dash: Accelerating Distributed Private Convolutional Neural Network Inference with Arithmetic Garbled Circuits