Paper 2020/396
Improving Non-Profiled Side-Channel Attacks using Autoencoder based Preprocessing
Donggeun Kwon and HeeSeok Kim and Seokhie Hong
Abstract
In recent years, deep learning-based side-channel attacks have established its position as mainstream. However, most deep learning techniques for cryptanalysis mainly focused on classifying side-channel information in a profiled scenario where attackers can obtain the label of training data. In this paper, we introduce a novel approach with deep learning for improving side-channel attacks, especially in a non-profiling scenario. We also propose a new method that trains autoencoder through the noise from real data using the noise-reduced-label. It notably diminishes the noise in a trace by adapting the autoencoder framework to the signal preprocessing. We show the convincing comparison from our custom dataset, which captured that our works outperform conventional preprocessing methods such as principal component analysis and linear discriminant analysis. Furthermore, we apply the method to realign desynchronized traces that applied hiding countermeasures, and we experimentally validate the performance of the proposal. Also, for masking countermeasures, we experimentally show that we can improve the performance of side-channel analysis by using an existing combining function or proposed method using domain knowledge.
Metadata
- Available format(s)
- Publication info
- Preprint. MINOR revision.
- Keywords
- Side-channel attackNon-profilingDeep learningAutoencoderPreprocessing
- Contact author(s)
- donggeun kwon @ gmail com
- History
- 2020-10-12: revised
- 2020-04-09: received
- See all versions
- Short URL
- https://ia.cr/2020/396
- License
-
CC BY