Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent

Thijs Veugen; Bart Kamphorst; Natasja van de L’Isle; Marie Beth van Egmond

doi:10.1007/978-3-030-78086-9_3

Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent

Thijs Veugen^*
, Bart Kamphorst
, Natasja van de L’Isle
, Marie Beth van Egmond

^*Corresponding author for this work

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

5 Citations (Scopus)

Abstract

We show how multiple data-owning parties can collaboratively train several machine learning algorithms without jeopardizing the privacy of their sensitive data. In particular, we assume that every party knows specific features of an overlapping set of people. Using a secure implementation of an advanced hidden set intersection protocol and a privacy-preserving Gradient Descent algorithm, we are able to train a Ridge, LASSO or SVM model over the intersection of people in their data sets. Both the hidden set intersection protocol and privacy-preserving LASSO implementation are unprecedented in literature.

Original language	English
Title of host publication	Cyber Security Cryptography and Machine Learning
Subtitle of host publication	5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings
Editors	Shlomi Dolev, Oded Margalit, Benny Pinkas, Alexander Schwarzmann
Place of Publication	Cham
Publisher	Springer
Pages	38-51
Number of pages	14
ISBN (Electronic)	978-3-030-78086-9
ISBN (Print)	978-3-030-78085-2
DOIs	https://doi.org/10.1007/978-3-030-78086-9_3
Publication status	Published - 2021
Externally published	Yes
Event	5th International Symposium on Cyber Security Cryptography and Machine Learning, CSCML 2021 - Be'er Sheva, Israel Duration: 8 Jul 2021 → 9 Jul 2021 Conference number: 5

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	12716
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

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

Keywords

Gradient descent
Privacy-preserving LASSO regression
Secure multi-party computation
Secure set intersection
n/a OA procedure

Access to Document

10.1007/978-3-030-78086-9_3

Cite this

Veugen, T., Kamphorst, B., van de L’Isle, N., & van Egmond, M. B. (2021). Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent. In S. Dolev, O. Margalit, B. Pinkas, & A. Schwarzmann (Eds.), Cyber Security Cryptography and Machine Learning: 5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings (pp. 38-51). (Lecture Notes in Computer Science; Vol. 12716). Springer. https://doi.org/10.1007/978-3-030-78086-9_3

Veugen, Thijs ; Kamphorst, Bart ; van de L’Isle, Natasja et al. / Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent. Cyber Security Cryptography and Machine Learning: 5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings. editor / Shlomi Dolev ; Oded Margalit ; Benny Pinkas ; Alexander Schwarzmann. Cham : Springer, 2021. pp. 38-51 (Lecture Notes in Computer Science).

@inproceedings{98fb5718bbf24e43a7b27894cb5994c7,

title = "Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent",

abstract = "We show how multiple data-owning parties can collaboratively train several machine learning algorithms without jeopardizing the privacy of their sensitive data. In particular, we assume that every party knows specific features of an overlapping set of people. Using a secure implementation of an advanced hidden set intersection protocol and a privacy-preserving Gradient Descent algorithm, we are able to train a Ridge, LASSO or SVM model over the intersection of people in their data sets. Both the hidden set intersection protocol and privacy-preserving LASSO implementation are unprecedented in literature.",

keywords = "Gradient descent, Privacy-preserving LASSO regression, Secure multi-party computation, Secure set intersection, n/a OA procedure",

author = "Thijs Veugen and Bart Kamphorst and \{van de L{\textquoteright}Isle\}, Natasja and \{van Egmond\}, \{Marie Beth\}",

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

year = "2021",

doi = "10.1007/978-3-030-78086-9\_3",

language = "English",

isbn = "978-3-030-78085-2",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "38--51",

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

booktitle = "Cyber Security Cryptography and Machine Learning",

address = "Germany",

}

Veugen, T, Kamphorst, B, van de L’Isle, N & van Egmond, MB 2021, Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent. in S Dolev, O Margalit, B Pinkas & A Schwarzmann (eds), Cyber Security Cryptography and Machine Learning: 5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings. Lecture Notes in Computer Science, vol. 12716, Springer, Cham, pp. 38-51, 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_3

Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent. / Veugen, Thijs; Kamphorst, Bart; van de L’Isle, Natasja et al.
Cyber Security Cryptography and Machine Learning: 5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings. ed. / Shlomi Dolev; Oded Margalit; Benny Pinkas; Alexander Schwarzmann. Cham: Springer, 2021. p. 38-51 (Lecture Notes in Computer Science; Vol. 12716).

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

TY - GEN

T1 - Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent

AU - Veugen, Thijs

AU - Kamphorst, Bart

AU - van de L’Isle, Natasja

AU - van Egmond, Marie Beth

N1 - Conference code: 5

PY - 2021

Y1 - 2021

N2 - We show how multiple data-owning parties can collaboratively train several machine learning algorithms without jeopardizing the privacy of their sensitive data. In particular, we assume that every party knows specific features of an overlapping set of people. Using a secure implementation of an advanced hidden set intersection protocol and a privacy-preserving Gradient Descent algorithm, we are able to train a Ridge, LASSO or SVM model over the intersection of people in their data sets. Both the hidden set intersection protocol and privacy-preserving LASSO implementation are unprecedented in literature.

AB - We show how multiple data-owning parties can collaboratively train several machine learning algorithms without jeopardizing the privacy of their sensitive data. In particular, we assume that every party knows specific features of an overlapping set of people. Using a secure implementation of an advanced hidden set intersection protocol and a privacy-preserving Gradient Descent algorithm, we are able to train a Ridge, LASSO or SVM model over the intersection of people in their data sets. Both the hidden set intersection protocol and privacy-preserving LASSO implementation are unprecedented in literature.

KW - Gradient descent

KW - Privacy-preserving LASSO regression

KW - Secure multi-party computation

KW - Secure set intersection

KW - n/a OA procedure

UR - https://www.scopus.com/pages/publications/85111959576

U2 - 10.1007/978-3-030-78086-9_3

DO - 10.1007/978-3-030-78086-9_3

M3 - Conference contribution

AN - SCOPUS:85111959576

SN - 978-3-030-78085-2

T3 - Lecture Notes in Computer Science

SP - 38

EP - 51

BT - Cyber Security Cryptography and Machine Learning

A2 - Dolev, Shlomi

A2 - Margalit, Oded

A2 - Pinkas, Benny

A2 - Schwarzmann, Alexander

PB - Springer

CY - Cham

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

Y2 - 8 July 2021 through 9 July 2021

ER -

Veugen T, Kamphorst B, van de L’Isle N, van Egmond MB. Privacy-Preserving Coupling of Vertically-Partitioned Databases and Subsequent Training with Gradient Descent. In Dolev S, Margalit O, Pinkas B, Schwarzmann A, editors, Cyber Security Cryptography and Machine Learning: 5th International Symposium, CSCML 2021, Be'er Sheva, Israel, July 8–9, 2021, Proceedings. Cham: Springer. 2021. p. 38-51. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-78086-9_3