Applications of machine learning techniques in side-channel attacks: a survey

With increasing expansion of the Internet of Things, embedded devices equipped with cryptographic modules become an important factor to protect sensitive data. Even though the employed algorithms in such devices are mathematically secure in theory, adversaries may still be able to compromise them by...

Full description

Saved in:
Bibliographic Details
Published in:Journal of cryptographic engineering 2020-06, Vol.10 (2), p.135-162
Main Authors: Hettwer, Benjamin, Gehrer, Stefan, Güneysu, Tim
Format: Article
Language:English
Subjects:
Algorithms
Circuits and Systems
Communications Engineering
Computer Communication Networks
Computer Science
Cryptography
Cryptology
Data Structures and Information Theory
Electronic devices
Embedded systems
Internet of Things
Machine learning
Networks
Operating Systems
Power consumption
Power management
Regular Paper
Workflow
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With increasing expansion of the Internet of Things, embedded devices equipped with cryptographic modules become an important factor to protect sensitive data. Even though the employed algorithms in such devices are mathematically secure in theory, adversaries may still be able to compromise them by means of side-channel attacks. In power-based side-channel attacks, the instantaneous power consumption of the target is analyzed with statistical tools to draw conclusions about the secret keys that are used. There is a recent line of work that additionally makes use of techniques from the machine learning domain to attack cryptographic implementations. Since a complete review of this emerging field has not been done so far, this research aims to survey the current state of the art. We use a target-based classification to differentiate published work and drive general conclusions according to a common machine learning workflow. Furthermore, we outline the relationship between traditional power analysis techniques and machine learning-based attacks. This enables researchers to gain a better understanding of the topic in order to design new attack methods as well as potential countermeasures.
ISSN:2190-8508
2190-8516
DOI:10.1007/s13389-019-00212-8