A Comparison of Multi-task Learning and Single-Task Learning Approaches

Thomas Marquet; Elisabeth Oswald

doi:10.1007/978-3-031-41181-6_7

A Comparison of Multi-task Learning and Single-Task Learning Approaches

Thomas Marquet^*, Elisabeth Oswald

^*Corresponding author for this work

Computer Science

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

Abstract

In this paper, we provide experimental evidence for the benefits of multi-task learning in the context of masked AES implementations (via the ASCADv1-r and ASCADv2 databases). We develop an approach for comparing single-task and multi-task approaches rather than comparing specific resulting models: we do this by training many models with random hyperparameters (instead of comparing a few highly tuned models). We find that multi-task learning has significant practical advantages that make it an attractive option in the context of device evaluations: the multi-task approach leads to performant networks quickly in particular in situations where knowledge of internal randomness is not available during training.

Original language	English
Title of host publication	Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings
Publisher	Springer
Pages	121-138
Number of pages	18
ISBN (Electronic)	9783031411816
ISBN (Print)	9783031411809
DOIs	https://doi.org/10.1007/978-3-031-41181-6_7
Publication status	Published - 4 Oct 2023
Event	21st International Conference on Applied Cryptography and Network Security, ACNS 2023 - Kyoto, Japan Duration: 19 Jun 2023 → 22 Jun 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13907
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	21st International Conference on Applied Cryptography and Network Security, ACNS 2023
Country/Territory	Japan
City	Kyoto
Period	19/06/23 → 22/06/23

Bibliographical note

Funding Information:
Thomas Marquet has been supported by the KWF under grant number KWF-3520-31870-45842. Elisabeth Oswald has been supported in part by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement No 725042).

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Deep Learning
Masking
Multi-Task Learning
Side Channel Attacks

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-41181-6_7Licence: None: All rights reserved

Cite this

Marquet, T., & Oswald, E. (2023). A Comparison of Multi-task Learning and Single-Task Learning Approaches. In Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings (pp. 121-138). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13907). Springer. https://doi.org/10.1007/978-3-031-41181-6_7

Marquet, Thomas ; Oswald, Elisabeth. / A Comparison of Multi-task Learning and Single-Task Learning Approaches. Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings. Springer, 2023. pp. 121-138 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f41a12de2ee94169946769e094fd7695,

title = "A Comparison of Multi-task Learning and Single-Task Learning Approaches",

abstract = "In this paper, we provide experimental evidence for the benefits of multi-task learning in the context of masked AES implementations (via the ASCADv1-r and ASCADv2 databases). We develop an approach for comparing single-task and multi-task approaches rather than comparing specific resulting models: we do this by training many models with random hyperparameters (instead of comparing a few highly tuned models). We find that multi-task learning has significant practical advantages that make it an attractive option in the context of device evaluations: the multi-task approach leads to performant networks quickly in particular in situations where knowledge of internal randomness is not available during training.",

keywords = "Deep Learning, Masking, Multi-Task Learning, Side Channel Attacks",

author = "Thomas Marquet and Elisabeth Oswald",

note = "Funding Information: Thomas Marquet has been supported by the KWF under grant number KWF-3520-31870-45842. Elisabeth Oswald has been supported in part by the European Research Council (ERC) under the European Union{\textquoteright}s Horizon 2020 research and innovation program (grant agreement No 725042). Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 21st International Conference on Applied Cryptography and Network Security, ACNS 2023 ; Conference date: 19-06-2023 Through 22-06-2023",

year = "2023",

month = oct,

day = "4",

doi = "10.1007/978-3-031-41181-6_7",

language = "English",

isbn = "9783031411809",

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

publisher = "Springer",

pages = "121--138",

booktitle = "Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings",

}

Marquet, T & Oswald, E 2023, A Comparison of Multi-task Learning and Single-Task Learning Approaches. in Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13907, Springer, pp. 121-138, 21st International Conference on Applied Cryptography and Network Security, ACNS 2023, Kyoto, Japan, 19/06/23. https://doi.org/10.1007/978-3-031-41181-6_7

A Comparison of Multi-task Learning and Single-Task Learning Approaches. / Marquet, Thomas; Oswald, Elisabeth.
Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings. Springer, 2023. p. 121-138 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13907).

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

TY - GEN

T1 - A Comparison of Multi-task Learning and Single-Task Learning Approaches

AU - Marquet, Thomas

AU - Oswald, Elisabeth

N1 - Funding Information: Thomas Marquet has been supported by the KWF under grant number KWF-3520-31870-45842. Elisabeth Oswald has been supported in part by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement No 725042). Publisher Copyright: © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2023/10/4

Y1 - 2023/10/4

N2 - In this paper, we provide experimental evidence for the benefits of multi-task learning in the context of masked AES implementations (via the ASCADv1-r and ASCADv2 databases). We develop an approach for comparing single-task and multi-task approaches rather than comparing specific resulting models: we do this by training many models with random hyperparameters (instead of comparing a few highly tuned models). We find that multi-task learning has significant practical advantages that make it an attractive option in the context of device evaluations: the multi-task approach leads to performant networks quickly in particular in situations where knowledge of internal randomness is not available during training.

AB - In this paper, we provide experimental evidence for the benefits of multi-task learning in the context of masked AES implementations (via the ASCADv1-r and ASCADv2 databases). We develop an approach for comparing single-task and multi-task approaches rather than comparing specific resulting models: we do this by training many models with random hyperparameters (instead of comparing a few highly tuned models). We find that multi-task learning has significant practical advantages that make it an attractive option in the context of device evaluations: the multi-task approach leads to performant networks quickly in particular in situations where knowledge of internal randomness is not available during training.

KW - Deep Learning

KW - Masking

KW - Multi-Task Learning

KW - Side Channel Attacks

UR - http://www.scopus.com/inward/record.url?scp=85174442980&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-41181-6_7

DO - 10.1007/978-3-031-41181-6_7

M3 - Conference contribution

AN - SCOPUS:85174442980

SN - 9783031411809

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

SP - 121

EP - 138

BT - Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings

PB - Springer

T2 - 21st International Conference on Applied Cryptography and Network Security, ACNS 2023

Y2 - 19 June 2023 through 22 June 2023

ER -

Marquet T, Oswald E. A Comparison of Multi-task Learning and Single-Task Learning Approaches. In Applied Cryptography and Network Security Workshops - ACNS 2023 Satellite Workshops, ADSC, AIBlock, AIHWS, AIoTS, CIMSS, Cloud S and P, SCI, SecMT, SiMLA, Proceedings. Springer. 2023. p. 121-138. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-41181-6_7