Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Glass stabilized ultra-stable dual-emitting Mn-doped cesium lead halide perovskite quantum dots for cryogenic temperature sensing.

Nanoscale·2019

Same author

Simultaneous Tailoring of Dual-Phase Fluoride Precipitation and Dopant Distribution in Glass to Control Upconverting Luminescence.

ACS applied materials & interfaces·2019

Same author

<i>Nkx2-5</i> defines a subpopulation of pacemaker cells and is essential for the physiological function of the sinoatrial node in mice.

Development (Cambridge, England)·2019

Same author

Process Analysis of Anaerobic Fermentation Exposure to Metal Mixtures.

International journal of environmental research and public health·2019

Same author

Surface-to-Bulk Redox Coupling through Thermally Driven Li Redistribution in Li- and Mn-Rich Layered Cathode Materials.

Journal of the American Chemical Society·2019

Same author

Chlorine-additive-promoted incorporation of Mn<sup>2+</sup> dopants into CsPbCl<sub>3</sub> perovskite nanocrystals.

Nanoscale·2019

Search research articles

Related Experiment Video

Updated: Jul 5, 2025

P300-Based Brain-Computer Interface Speller Performance Estimation with Classifier-Based Latency Estimation

P300-Based Brain-Computer Interface Speller Performance Estimation with Classifier-Based Latency Estimation

Published on: September 8, 2023

SecBERT: Privacy-preserving pre-training based neural network inference system.

Hai Huang¹, Yongjian Wang¹

¹Computer School, Zhejiang Sci-Tech University, Hangzhou, 310018, China.

Neural Networks : the Official Journal of the International Neural Network Society

|January 25, 2024

Summary

This summary is machine-generated.

This study introduces SecBERT, a novel privacy-preserving method for neural network inference using BERT. It enables secure natural language processing tasks between a client and two servers without data disclosure.

Keywords:

BERT Neural network inference Pre-trained model Privacy-preserving computation Secret sharing

More Related Videos

A Fully Automated and Highly Versatile System for Testing Multi-cognitive Functions and Recording Neuronal Activities in Rodents

A Fully Automated and Highly Versatile System for Testing Multi-cognitive Functions and Recording Neuronal Activities in Rodents

Published on: May 3, 2012

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Related Experiment Videos

Last Updated: Jul 5, 2025

P300-Based Brain-Computer Interface Speller Performance Estimation with Classifier-Based Latency Estimation

P300-Based Brain-Computer Interface Speller Performance Estimation with Classifier-Based Latency Estimation

Published on: September 8, 2023

A Fully Automated and Highly Versatile System for Testing Multi-cognitive Functions and Recording Neuronal Activities in Rodents

A Fully Automated and Highly Versatile System for Testing Multi-cognitive Functions and Recording Neuronal Activities in Rodents

Published on: May 3, 2012

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Area of Science:

Computer Science
Cryptography
Artificial Intelligence

Background:

Pre-trained models like BERT excel in natural language processing (NLP).
Existing methods often lack robust privacy guarantees for sensitive data during inference.
Secure computation is crucial for privacy-preserving AI applications.

Purpose of the Study:

To develop a cryptographically secure, privacy-preserving inference protocol for pre-trained neural networks.
To enable resource-constrained clients to leverage powerful servers for NLP tasks without compromising data privacy.
To design secure sub-protocols for non-linear functions essential to BERT.

Main Methods:

Utilized additive secret sharing for a two-server framework.
Designed secure sub-protocols for non-linear functions within BERT.
Developed SecBERT, a novel privacy-preserving inference protocol.

Main Results:

SecBERT provides the first cryptographically secure protocol for privacy-preserving pre-trained neural network inference.
Demonstrated the security, efficiency, and accuracy of the SecBERT protocol through theoretical analysis and experiments.
Developed reusable secure sub-protocols for common non-linear functions.

Conclusions:

SecBERT offers a viable solution for privacy-preserving NLP inference.
The proposed sub-protocols have potential applications beyond BERT.
This work advances secure AI by enabling private computation on sensitive data.