Abstract
Recently, many studies have shown that using deep learning for side-channel attacks offers several advantages, including simplification of the attack phase and target breaking, even in protected implementations, while presenting outstanding attack performance. Power and electromagnetic analysis, which is known as the most robust attack, can be classified into profiling and non-profiling attacks. In the real world, a non-profiling attack is more ideal than a profiling attack. In particular, studies on non-profiling attacks using deep learning for asymmetric cryptosystems are rare and have shortcomings, such as a long analysis time. In this paper, we propose a novel non-profiling attack method for asymmetric cryptosystems that requires only a single trace and a reasonably short attack time to recover a full private key, overcoming the limitations of previous studies. The proposed method applies one-shot learning with a convolutional Siamese network, which is used for the first time in side-channel attacks. Thus, our proposed method can leak private keys used in a protected public-key cryptosystem with up to 100% accuracy with only one single trace in a non-profiled setting.
Original language | English |
---|---|
Pages (from-to) | 60778-60789 |
Number of pages | 12 |
Journal | IEEE Access |
Volume | 10 |
DOIs | |
Publication status | Published - 2022 |
Bibliographical note
Funding Information:This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant NRF-2019R1A2C2088960.
Publisher Copyright:
© 2013 IEEE.
Keywords
- Deep learning
- ECC
- Montgomery ladder
- Siamese network
- non-profiling attack
- one-shot learning
- side channel attack
- similarity score
ASJC Scopus subject areas
- General Computer Science
- General Materials Science
- General Engineering
- Electrical and Electronic Engineering