Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption: Opportunities and Limitations

Fabian Schmid; Christian Rechberger; Shibam Mukherjee; Marc Stöttinger; Stjepan Picek; Fabrizio De Santis

Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption: Opportunities and Limitations

Titel in Übersetzung: Private Deep Learning-basierte Seitenkanalanalyse mit homomorpher Verschlüsselung: Möglichkeiten und Grenzen

Fabian Schmid^*, Christian Rechberger, Shibam Mukherjee, Marc Stöttinger, Stjepan Picek, Fabrizio De Santis

^*Korrespondierende/r Autor/-in für diese Arbeit

Institut für Angewandte Informationsverarbeitung und Kommunikationstechnologie (7050)

Publikation: Beitrag in Buch/Bericht/Konferenzband › Beitrag in einem Konferenzband › Begutachtung

Abstract

In dieser Arbeit wird die Verwendung homomorpher Verschlüsselung (HE) zur Unterstützung der Sicherheitsevaluierung von kryptografischen Geräten untersucht, ohne dass Seitenkanalinformationen offengelegt werden. Zum ersten Mal evaluieren wir die Machbarkeit der Ausführung von Deep-Learning-basierten Seitenkanalanalysen auf Standard-Servergeräten unter Verwendung eines angepassten HE-Protokolls. Durch die Untersuchung von Genauigkeit und Ausführungszeit wird die erfolgreiche Anwendung von privater SCA sowohl auf ungeschützte als auch auf geschützte kryptografische Implementierungen demonstriert. Unsere Studie beschränkt sich auf das Modell des "honest-but-curious adversaries", bei dem wir das Geheimnis einer ungeschützten AES-Implementierung in Sekunden und das einer maskierten AES-Implementierung in weniger als 17 Minuten rekonstruieren konnten.

Titel in Übersetzung	Private Deep Learning-basierte Seitenkanalanalyse mit homomorpher Verschlüsselung: Möglichkeiten und Grenzen
Originalsprache	englisch
Titel	Constructive Side-Channel Analysis and Secure Design
Herausgeber (Verlag)	Springer
Publikationsstatus	Angenommen/In Druck - 31 Jan. 2024
Veranstaltung	15th International Workshop on Constructive Side-channel Analysis and Secure Design: COSADE 2024 - 880, route de Mimet, Gardanne, Frankreich Dauer: 8 Apr. 2024 → 10 Apr. 2024 https://www.cosade.org/cosade24/program.html

Konferenz

Konferenz	15th International Workshop on Constructive Side-channel Analysis and Secure Design
Kurztitel	COSADE
Land/Gebiet	Frankreich
Ort	Gardanne
Zeitraum	8/04/24 → 10/04/24
Internetadresse	https://www.cosade.org/cosade24/program.html

Schlagwörter

Homomorphic Encryption
Neural Networks
Side-Channel Analysis

COSADE 2024
Fabian Schmid (Redner/in)
10 Apr. 2024
Aktivität: Vortrag oder Präsentation › Vortrag bei Konferenz oder Fachtagung › Science to science

Dieses zitieren

Schmid, F , Rechberger, C , Mukherjee, S, Stöttinger, M, Picek, S & De Santis, F 2024, Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption: Opportunities and Limitations. in Constructive Side-Channel Analysis and Secure Design . Springer, 15th International Workshop on Constructive Side-channel Analysis and Secure Design, Gardanne, Frankreich, 8/04/24.

@inproceedings{1f5e62f8b2cb407e9e75d28134276179,

title = "Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption: Opportunities and Limitations",

abstract = "This work investigates using Homomorphic Encryption (HE) to assist the security evaluation of cryptographic devices without revealing side-channel information. For the first time, we evaluate the feasibility of executing deep learning-based side-channel analysis on standard server equipment using an adapted HE protocol. By examining accuracy and execution time, it demonstrates the successful application of private SCA on both unprotected and protected cryptographic implementations. This contribution is a first step towards confidential side-channel analysis. Our study is limited to the honest-but-curious trust model, where we could reconstruct the secret of an unprotected AES implementation in seconds and of a masked AES implementation in under 17 minutes.",

keywords = "Homomorphic Encryption, Neural Networks, Side-Channel Analysis, Homomorphic Encryption, Neural Networks, Side-Channel Analysis",

author = "Fabian Schmid and Christian Rechberger and Shibam Mukherjee and Marc St{\"o}ttinger and Stjepan Picek and {De Santis}, Fabrizio",

year = "2024",

month = jan,

day = "31",

language = "English",

booktitle = "Constructive Side-Channel Analysis and Secure Design",

publisher = "Springer",

note = "15th International Workshop on Constructive Side-Channel Analysis and Secure Design : COSADE 2024, COSADE 2024 ; Conference date: 08-04-2024 Through 10-04-2024",

url = "https://www.cosade.org/cosade24/program.html",

}

TY - GEN

T1 - Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption

T2 - 15th International Workshop on Constructive Side-Channel Analysis and Secure Design

AU - Schmid, Fabian

AU - Rechberger, Christian

AU - Mukherjee, Shibam

AU - Stöttinger, Marc

AU - Picek, Stjepan

AU - De Santis, Fabrizio

PY - 2024/1/31

Y1 - 2024/1/31

N2 - This work investigates using Homomorphic Encryption (HE) to assist the security evaluation of cryptographic devices without revealing side-channel information. For the first time, we evaluate the feasibility of executing deep learning-based side-channel analysis on standard server equipment using an adapted HE protocol. By examining accuracy and execution time, it demonstrates the successful application of private SCA on both unprotected and protected cryptographic implementations. This contribution is a first step towards confidential side-channel analysis. Our study is limited to the honest-but-curious trust model, where we could reconstruct the secret of an unprotected AES implementation in seconds and of a masked AES implementation in under 17 minutes.

AB - This work investigates using Homomorphic Encryption (HE) to assist the security evaluation of cryptographic devices without revealing side-channel information. For the first time, we evaluate the feasibility of executing deep learning-based side-channel analysis on standard server equipment using an adapted HE protocol. By examining accuracy and execution time, it demonstrates the successful application of private SCA on both unprotected and protected cryptographic implementations. This contribution is a first step towards confidential side-channel analysis. Our study is limited to the honest-but-curious trust model, where we could reconstruct the secret of an unprotected AES implementation in seconds and of a masked AES implementation in under 17 minutes.

KW - Homomorphic Encryption

KW - Neural Networks

KW - Side-Channel Analysis

KW - Homomorphic Encryption

KW - Neural Networks

KW - Side-Channel Analysis

M3 - Conference paper

BT - Constructive Side-Channel Analysis and Secure Design

PB - Springer

Y2 - 8 April 2024 through 10 April 2024

ER -

Towards Private Deep Learning-Based Side-Channel Analysis using Homomorphic Encryption: Opportunities and Limitations

Abstract

Konferenz

Schlagwörter

Fingerprint

Aktivitäten

COSADE 2024

Dieses zitieren