Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks

Md Shohidul Islam; Ihsen Alouani; Khaled N. Khasawneh

doi:10.1109/ICCAD51958.2021.9643551

Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks

Md Shohidul Islam, Ihsen Alouani, Khaled N. Khasawneh

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

8 Citations (Scopus)

97 Downloads (Pure)

Abstract

Deep neural networks (DNNs) are shown to be vulnerable to adversarial attacks—– carefully crafted additive noise that undermines DNNs integrity. Previously proposed defenses against these attacks require substantial overheads, making it challenging to deploy these solutions in power and computational resource-constrained devices, such as embedded systems and the Edge. In this paper, we explore the use of voltage overscaling (VOS) as a lightweight defense against adversarial attacks. Specifically, we exploit the stochastic timing violations of VOS to implement a moving-target defense for DNNs. Our experimental results demonstrate that VOS guarantees effective defense against different attack methods, does not require any software/hardware modifications, and offers a by-product reduction in power consumption.

Original language	English
Title of host publication	2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781665445078
DOIs	https://doi.org/10.1109/ICCAD51958.2021.9643551
Publication status	Published - 23 Dec 2021
Externally published	Yes
Event	40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Munich, Germany Duration: 01 Nov 2021 → 04 Nov 2021

Publication series

Name	IEEE/ACM International Conference on Computer-Aided Design: Proceedings
Publisher	IEEE
ISSN (Print)	1933-7760
ISSN (Electronic)	1558-2434

Conference

Conference	40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021
Country/Territory	Germany
City	Munich
Period	01/11/2021 → 04/11/2021

Bibliographical note

Publisher Copyright:
© 2021 IEEE.

Keywords

Adversarial attacks
Approximate computing
Machine learning
Voltage overscaling

ASJC Scopus subject areas

Software
Computer Science Applications
Computer Graphics and Computer-Aided Design

Access to Document

10.1109/ICCAD51958.2021.9643551Licence: Unspecified

Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks
© 2022 IEEE/ACM This work is made available online in accordance with the publisher’s policies. Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 508 KBLicence: Unspecified

Cite this

Islam, M. S., Alouani, I., & Khasawneh, K. N. (2021). Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks. In 2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings (IEEE/ACM International Conference on Computer-Aided Design: Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICCAD51958.2021.9643551

Islam, Md Shohidul ; Alouani, Ihsen ; Khasawneh, Khaled N. / Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks. 2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2021. (IEEE/ACM International Conference on Computer-Aided Design: Proceedings).

@inproceedings{cecc5a117afe468192ba8eb10430700b,

title = "Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks",

abstract = "Deep neural networks (DNNs) are shown to be vulnerable to adversarial attacks—– carefully crafted additive noise that undermines DNNs integrity. Previously proposed defenses against these attacks require substantial overheads, making it challenging to deploy these solutions in power and computational resource-constrained devices, such as embedded systems and the Edge. In this paper, we explore the use of voltage overscaling (VOS) as a lightweight defense against adversarial attacks. Specifically, we exploit the stochastic timing violations of VOS to implement a moving-target defense for DNNs. Our experimental results demonstrate that VOS guarantees effective defense against different attack methods, does not require any software/hardware modifications, and offers a by-product reduction in power consumption.",

keywords = "Adversarial attacks, Approximate computing, Machine learning, Voltage overscaling",

author = "Islam, {Md Shohidul} and Ihsen Alouani and Khasawneh, {Khaled N.}",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 ; Conference date: 01-11-2021 Through 04-11-2021",

year = "2021",

month = dec,

day = "23",

doi = "10.1109/ICCAD51958.2021.9643551",

language = "English",

series = "IEEE/ACM International Conference on Computer-Aided Design: Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings",

address = "United States",

}

Islam, MS, Alouani, I & Khasawneh, KN 2021, Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks. in 2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings. IEEE/ACM International Conference on Computer-Aided Design: Proceedings, Institute of Electrical and Electronics Engineers Inc., 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021, Munich, Germany, 01/11/2021. https://doi.org/10.1109/ICCAD51958.2021.9643551

Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks. / Islam, Md Shohidul; Alouani, Ihsen; Khasawneh, Khaled N.
2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2021. (IEEE/ACM International Conference on Computer-Aided Design: Proceedings).

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

TY - GEN

T1 - Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks

AU - Islam, Md Shohidul

AU - Alouani, Ihsen

AU - Khasawneh, Khaled N.

PY - 2021/12/23

Y1 - 2021/12/23

N2 - Deep neural networks (DNNs) are shown to be vulnerable to adversarial attacks—– carefully crafted additive noise that undermines DNNs integrity. Previously proposed defenses against these attacks require substantial overheads, making it challenging to deploy these solutions in power and computational resource-constrained devices, such as embedded systems and the Edge. In this paper, we explore the use of voltage overscaling (VOS) as a lightweight defense against adversarial attacks. Specifically, we exploit the stochastic timing violations of VOS to implement a moving-target defense for DNNs. Our experimental results demonstrate that VOS guarantees effective defense against different attack methods, does not require any software/hardware modifications, and offers a by-product reduction in power consumption.

AB - Deep neural networks (DNNs) are shown to be vulnerable to adversarial attacks—– carefully crafted additive noise that undermines DNNs integrity. Previously proposed defenses against these attacks require substantial overheads, making it challenging to deploy these solutions in power and computational resource-constrained devices, such as embedded systems and the Edge. In this paper, we explore the use of voltage overscaling (VOS) as a lightweight defense against adversarial attacks. Specifically, we exploit the stochastic timing violations of VOS to implement a moving-target defense for DNNs. Our experimental results demonstrate that VOS guarantees effective defense against different attack methods, does not require any software/hardware modifications, and offers a by-product reduction in power consumption.

KW - Adversarial attacks

KW - Approximate computing

KW - Machine learning

KW - Voltage overscaling

UR - http://www.scopus.com/inward/record.url?scp=85124129642&partnerID=8YFLogxK

U2 - 10.1109/ICCAD51958.2021.9643551

DO - 10.1109/ICCAD51958.2021.9643551

M3 - Conference contribution

AN - SCOPUS:85124129642

T3 - IEEE/ACM International Conference on Computer-Aided Design: Proceedings

BT - 2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021

Y2 - 1 November 2021 through 4 November 2021

ER -

Islam MS, Alouani I, Khasawneh KN. Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks. In 2021 40th IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2021 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2021. (IEEE/ACM International Conference on Computer-Aided Design: Proceedings). doi: 10.1109/ICCAD51958.2021.9643551

Lower Voltage for Higher Security: Using Voltage Overscaling to Secure Deep Neural Networks

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this