< Terug naar vorige pagina

Publicatie

Revisiting a Methodology for Efficient CNN Architectures in Profiling Attacks.

Tijdschriftbijdrage - Tijdschriftartikel

This work provides a critical review of the paper by Zaid et al. titled“Methodology for Efficient CNN Architectures in Profiling attacks”, which was pub-lished in TCHES Volume 2020, Issue 1. This work studies the design of CNN networksto perform side-channel analysis of multiple implementations of the AES for embeddeddevices. Based on the authors’ code and public data sets, we were able to cross-checktheir results and perform a thorough analysis. We correct multiple misconceptionsby carefully inspecting different elements of the model architectures proposed byZaid et al. First, by providing a better understanding on the internal workings ofthese models, we can trivially reduce their number of parameters on average by52%, while maintaining a similar performance. Second, we demonstrate that theconvolutional filter’s size isnotstrictly related to the amount of misalignment in thetraces. Third, we show that increasing the filter size and the number of convolutionsactuallyimprovesthe performance of a network. Our work demonstrates once againthat reproducibility and review are important pillars of academic research. Therefore,we provide the reader with an online Python notebook which allows to reproducesome of our experiments1and additional example code is made available on Github.
Tijdschrift: Transactions on Cryptographic Hardware and Embedded Systems
ISSN: 2569-2925
Issue: 3
Volume: 2020
Pagina's: 147 - 168
Jaar van publicatie:2020
Toegankelijkheid:Open