Publication detail

Profiling Power Analysis Attack Based on Multi-Layer Perceptron Network

MARTINÁSEK, Z. MALINA, L. TRÁSY, K.

Original Title

Profiling Power Analysis Attack Based on Multi-Layer Perceptron Network

Czech Title

Profilující útok proudovou analyzou využívající MLP

English Title

Profiling Power Analysis Attack Based on Multi-Layer Perceptron Network

Type

book chapter

Language

en

Original Abstract

In 2013, an innovative method of power analysis was presented. Realized experiments proved that the proposed method based on Multi-Layer Perceptron (MLP) can provide almost 100 percent success rate. This description based on the first-order success rate is not appropriate enough. Moreover, the above mentioned works contain other lacks: the MLP has not been compared with other well-known attacks, an adversary uses too many points of power trace and a general description of the MLP method was not provided. In this paper, we eliminate these weaknesses by introducing the first fair comparison of power analysis attacks based on the MLP and templates. The comparison is accomplished by using the identical data sets, number of interesting points and guessing entropy as a metric. The first data set created contains the power traces of an unprotected AES implementation in order to classify the secret key stored. The second and third data sets were created independently from public available power traces corresponding to a masked AES implementation (DPA Contest v4). Secret offset is revealed depending on the number of interesting points and power traces in this experiment. Moreover, we create a general description of the MLP attack.

Czech abstract

Proudová analýza představuje efektivní způsob útoku na kryptgrafické moduly. V dneštní době se používají algoritmy strojového učení namísto klasických šablon. V roce 2013 byla představena metoda využívajíc MLP. Články popisující tuto metodu obsahovaly nedostatky (chybně zvolen počet bodů, chybějící obecnéschéma). Článek eliminuje tyto nedostatky a popisuje obecné schéma útoku založeného na MLP. Dále je navrhovaná metoda porovnána s TA útoky na datových setech z DPA Contest v4.

English abstract

In 2013, an innovative method of power analysis was presented. Realized experiments proved that the proposed method based on Multi-Layer Perceptron (MLP) can provide almost 100 percent success rate. This description based on the first-order success rate is not appropriate enough. Moreover, the above mentioned works contain other lacks: the MLP has not been compared with other well-known attacks, an adversary uses too many points of power trace and a general description of the MLP method was not provided. In this paper, we eliminate these weaknesses by introducing the first fair comparison of power analysis attacks based on the MLP and templates. The comparison is accomplished by using the identical data sets, number of interesting points and guessing entropy as a metric. The first data set created contains the power traces of an unprotected AES implementation in order to classify the secret key stored. The second and third data sets were created independently from public available power traces corresponding to a masked AES implementation (DPA Contest v4). Secret offset is revealed depending on the number of interesting points and power traces in this experiment. Moreover, we create a general description of the MLP attack.

Keywords

Power Analysis, MLP, Machine Learning, Template Attack, Comparison.

RIV year

2015

Released

16.10.2015

Publisher

Springer International Publishing

Location

Switzerland 2015

ISBN

978-3-319-15764-1

Book

Computational Problems in Science and Engineering

Edition

1

Edition number

1

Pages from

1

Pages to

25

Pages count

25

BibTex


@inbook{BUT117847,
  author="Zdeněk {Martinásek} and Lukáš {Malina} and Krisztina {Trásy}",
  title="Profiling Power Analysis Attack Based on Multi-Layer Perceptron Network",
  annote="In 2013, an innovative method of power analysis was presented.
Realized experiments proved that the proposed method based on Multi-Layer Perceptron (MLP) can provide almost 100 percent success rate.
 This description based on the first-order success rate is not appropriate enough.
 Moreover, the above mentioned works contain other lacks: the MLP has not been compared with other well-known attacks, an adversary uses too many points of power trace and a general description of the MLP method was not provided. 
In this paper, we eliminate these weaknesses by introducing the first fair comparison of power analysis attacks based on the MLP and templates.
 The comparison is accomplished by using the identical data sets, number of interesting points and guessing entropy as a metric. The first data set created contains the power traces of an unprotected AES implementation in order to classify the secret key stored. 
The second and third data sets were created independently from public available power traces corresponding to a masked AES implementation (DPA Contest v4). 
Secret offset is revealed depending on the number of interesting points and power traces in this experiment. 
Moreover, we create a general description of the MLP attack.",
  address="Springer International Publishing",
  booktitle="Computational Problems in Science and Engineering",
  chapter="117847",
  doi="10.1007/978-3-319-15765-8",
  edition="1",
  howpublished="print",
  institution="Springer International Publishing",
  year="2015",
  month="october",
  pages="1--25",
  publisher="Springer International Publishing",
  type="book chapter"
}