Paper 2026/2904

Blind Deep-Learning Side-Channel Analysis of XTS-AES on a Commercial RISC-V Microcontroller

Keerthi Kamalakshan, Nanyang Technological University
Dirmanto Jap, Nanyang Technological University
Trevor Yap, Nanyang Technological University
Shivam Bhasin, Nanyang Technological University
Abstract

The XTS-AES mode is widely deployed to provide storage encryption in embedded systems, where hardware accelerators are commonly used to protect sensitive firmware stored in external flash. Existing side-channel attacks against XTS-AES typically assume access to operation-level leakage, requiring precise segmentation of individual AES executions or exploiting implementation-specific temporal structures. Such assumptions are difficult to satisfy for commercial hardware accelerators, where cryptographic operations are tightly integrated and exposed only through coarse-grained leakage. This paper presents a blind side-channel analysis methodology for hardware XTS-AES implementations that operates directly on complete power traces without requiring operation-level segmentation or knowledge of execution boundaries. The proposed approach adapted the blind deep learning-based side channel analysis to recover the XTS-AES encryption key from full-trace measurements. We further show that recovery of key can help to reconstruct the sector-specific XTS tweak, eliminating the need to recover the second XTS key. The proposed methodology is evaluated on the hardware XTS-AES of a commercial ESP32-C3 microcontroller. Experimental results demonstrate successful recovery of the encryption key, reconstruction of the sector-specific tweak, and demonstrates how this can be used to recover the flash contents. The results highlight the practical feasibility of blind whole-trace side-channel analysis against the security of hardware-accelerated XTS-AES designs.

Metadata
Available format(s)
PDF
Category
Attacks and cryptanalysis
Publication info
Preprint.
Keywords
Blind side-channel analysisFlash EncryptionXTS-AESAES acceleratorEmbedded Security.
Contact author(s)
keerthi kamalakshan @ ntu edu sg
djap @ ntu edu sg
trevor yap @ ntu edu sg
sbhasin @ ntu edu sg
History
2026-11-21: approved
2026-11-16: received
See all versions
Short URL
https://ia.cr/2026/2904
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/2904,
      author = {Keerthi Kamalakshan and Dirmanto Jap and Trevor Yap and Shivam Bhasin},
      title = {Blind Deep-Learning Side-Channel Analysis of {XTS}-{AES} on a Commercial {RISC}-V Microcontroller},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/2904},
      year = {2026},
      url = {https://eprint.iacr.org/2026/2904}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.