PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference

Philip Derbeko; Shlomi Dolev

doi:10.1007/978-3-030-78086-9_24

PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference

Philip Derbeko, Shlomi Dolev

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

The structure and weights of Deep Neural Networks (DNN) typically encode and contain very valuable information about the dataset that was used to train the network. One way to protect this information when DNN is published is to perform an interference of the network using secure multi-party computations (MPC). In this paper, we suggest a translation of deep neural networks to polynomials, which are easier to calculate efficiently with MPC techniques. We show a way to translate complete networks into a single polynomial and how to calculate the polynomial with an efficient and information-secure MPC algorithm. The calculation is done without intermediate communication between the participating parties, which is beneficial in several cases, as explained in the paper.

Original language	English GB
Title of host publication	Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings
Editors	Shlomi Dolev, Oded Margalit, Benny Pinkas, Alexander Schwarzmann
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	317-324
Number of pages	8
ISBN (Print)	9783030780852
DOIs	https://doi.org/10.1007/978-3-030-78086-9_24
State	Published - 1 Jan 2021
Event	5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021 - Be'er Sheva, Israel Duration: 8 Jul 2021 → 9 Jul 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12716 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021
Country/Territory	Israel
City	Be'er Sheva
Period	8/07/21 → 9/07/21

Keywords

Data publishing
Data sharing
DNN
Privacy

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science (all)

Access to Document

10.1007/978-3-030-78086-9_24

Cite this

Derbeko, P., & Dolev, S. (2021). PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference. In S. Dolev, O. Margalit, B. Pinkas, & A. Schwarzmann (Eds.), Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings (pp. 317-324). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12716 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-78086-9_24

Derbeko, Philip ; Dolev, Shlomi. / PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference. Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings. editor / Shlomi Dolev ; Oded Margalit ; Benny Pinkas ; Alexander Schwarzmann. Springer Science and Business Media Deutschland GmbH, 2021. pp. 317-324 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{9344927ebdb948978878ca34b5af94ba,

title = "PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference",

abstract = "The structure and weights of Deep Neural Networks (DNN) typically encode and contain very valuable information about the dataset that was used to train the network. One way to protect this information when DNN is published is to perform an interference of the network using secure multi-party computations (MPC). In this paper, we suggest a translation of deep neural networks to polynomials, which are easier to calculate efficiently with MPC techniques. We show a way to translate complete networks into a single polynomial and how to calculate the polynomial with an efficient and information-secure MPC algorithm. The calculation is done without intermediate communication between the participating parties, which is beneficial in several cases, as explained in the paper.",

keywords = "Data publishing, Data sharing, DNN, Privacy",

author = "Philip Derbeko and Shlomi Dolev",

note = "Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021 ; Conference date: 08-07-2021 Through 09-07-2021",

year = "2021",

month = jan,

day = "1",

doi = "10.1007/978-3-030-78086-9_24",

language = "אנגלית",

isbn = "9783030780852",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "317--324",

editor = "Shlomi Dolev and Oded Margalit and Benny Pinkas and Alexander Schwarzmann",

booktitle = "Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings",

address = "גרמניה",

}

Derbeko, P & Dolev, S 2021, PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference. in S Dolev, O Margalit, B Pinkas & A Schwarzmann (eds), Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12716 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 317-324, 5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021, Be'er Sheva, Israel, 8/07/21. https://doi.org/10.1007/978-3-030-78086-9_24

PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference. / Derbeko, Philip; Dolev, Shlomi.
Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings. ed. / Shlomi Dolev; Oded Margalit; Benny Pinkas; Alexander Schwarzmann. Springer Science and Business Media Deutschland GmbH, 2021. p. 317-324 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12716 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference

AU - Derbeko, Philip

AU - Dolev, Shlomi

PY - 2021/1/1

Y1 - 2021/1/1

N2 - The structure and weights of Deep Neural Networks (DNN) typically encode and contain very valuable information about the dataset that was used to train the network. One way to protect this information when DNN is published is to perform an interference of the network using secure multi-party computations (MPC). In this paper, we suggest a translation of deep neural networks to polynomials, which are easier to calculate efficiently with MPC techniques. We show a way to translate complete networks into a single polynomial and how to calculate the polynomial with an efficient and information-secure MPC algorithm. The calculation is done without intermediate communication between the participating parties, which is beneficial in several cases, as explained in the paper.

AB - The structure and weights of Deep Neural Networks (DNN) typically encode and contain very valuable information about the dataset that was used to train the network. One way to protect this information when DNN is published is to perform an interference of the network using secure multi-party computations (MPC). In this paper, we suggest a translation of deep neural networks to polynomials, which are easier to calculate efficiently with MPC techniques. We show a way to translate complete networks into a single polynomial and how to calculate the polynomial with an efficient and information-secure MPC algorithm. The calculation is done without intermediate communication between the participating parties, which is beneficial in several cases, as explained in the paper.

KW - Data publishing

KW - Data sharing

KW - DNN

KW - Privacy

UR - http://www.scopus.com/inward/record.url?scp=85111953367&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-78086-9_24

DO - 10.1007/978-3-030-78086-9_24

M3 - פרסום בספר כנס

AN - SCOPUS:85111953367

SN - 9783030780852

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 317

EP - 324

BT - Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings

A2 - Dolev, Shlomi

A2 - Margalit, Oded

A2 - Pinkas, Benny

A2 - Schwarzmann, Alexander

PB - Springer Science and Business Media Deutschland GmbH

T2 - 5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021

Y2 - 8 July 2021 through 9 July 2021

ER -

Derbeko P, Dolev S. PolyDNN Polynomial Representation of NN for Communication-Less SMPC Inference. In Dolev S, Margalit O, Pinkas B, Schwarzmann A, editors, Cyber Security Cryptography and Machine Learning - 5th International Symposium, CSCML 2021, Proceedings. Springer Science and Business Media Deutschland GmbH. 2021. p. 317-324. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-78086-9_24