|
A Comprehensive Exploration of Deep Learning Approaches in Differential Cryptanalysis of Lightweight Block Ciphers
|
Iman Mirzaali Mazandarani1   , Nasour Bagheri *2 , Sadegh Sadeghi3 |
1- Department of TeleCommunications, Shahid Rajaee Teacher Training University, Tehran, IranShahid Rajaee Teacher Training University
2- Department of TeleCommunications, Shahid Rajaee Teacher Training University, Tehran, Iran
3- Department of Mathematics, Institute for Advanced Studies in Basic Sciences, Zanjan, Iran |
|
|
Abstract: (4206 Views) |
| With the increasing and widespread application of deep learning and neural networks across various scientific domains and the notable successes achieved, deep neural networks were employed for differential cryptanalysis in 2019. This marked the initiation of growing interest in this research domain. While most existing works primarily focus on enhancing and deploying neural distinguishers, limited studies have delved into the intrinsic principles and learned characteristics of these neural distinguishers. In this study, our focus will be on analyzing block ciphers such as Speck, Simon, and Simeck using deep learning. We will explore and compare the factors and components that contribute to better performance. Additionally, by detailing attacks and comparing results, we aim to address the question of whether neural networks and deep learning can effectively serve as tools for block cipher cryptanalysis or not. |
|
| Keywords: Block cipher, Cryptanalysis, Neural distinguisher, Key recovery, Deep learning, Neural network. |
|
|
Full-Text [PDF 2276 kb]
(2663 Downloads)
|
Type of Study: Review Article |
Subject:
Cryptology and Information Security
Received: 2023/12/15 | Accepted: 2024/02/17 | Published: 2024/03/26
|
|
|
|
|
|
