Paper 2026/2933

Hardware/Software Co-Design of ML-DSA with Random Delay Insertion Against Correlation Power Analysis

Salvatore Coppa, University of Salerno
Ettore Napoli, University of Salerno
Abstract

Post-Quantum Cryptography (PQC) recently led to the standardization of the Module-Lattice-Based Digital Signature Algorithm (ML-DSA). Standard software implementations of ML-DSA on embedded devices are exposed to physical side-channel attacks, since the power consumption of the circuit depends on the processed secret data and allows key recovery via Correlation Power Analysis (CPA). In addition, protecting these implementations in software introduces large latency overheads. In this paper, the physical vulnerability of ML-DSA is considered focusing the study on embedded microcontroller and System-on-Chip (SoC) platforms. The computational cost of software countermeasures, validated via Test Vector Leakage Assessment (TVLA), is evaluated on both architectures. An Hardware/Software co-design implemented on a Zynq-7000 SoC is then proposed to overcome the software execution bottleneck. The vulnerable pointwise Montgomery multiplication is offloaded to an Advanced eXtensible Interface (AXI4-Full) hardware accelerator that integrates a 16-bit Linear Feedback Shift Register (LFSR) for Random Delay Insertion (RDI), preventing CPA trace alignment. Experimental results demonstrate a 13.4× speedup over the software-protected SoC implementation. The proposed architecture reduces the protected signing latency to 42.80 ms, achieving an execution time close to the 40.64 ms of the unprotected software implementation, while the dynamic power dissipated by the hardware accelerator in the programmable logic is limited to 33 mW.

Metadata
Available format(s)
PDF
Category
Implementation
Publication info
Preprint.
Keywords
PQCML-DSASide-Channel AnalysisCorrelation Power AnalysisHW/SW Co-DesignRandom Delay Insertion
Contact author(s)
salvatorecoppa2002 @ gmail com
enapoli @ unisa it
History
2026-11-22: approved
2026-11-18: received
See all versions
Short URL
https://ia.cr/2026/2933
License
Creative Commons Attribution-NonCommercial-NoDerivs
CC BY-NC-ND

BibTeX

@misc{cryptoeprint:2026/2933,
      author = {Salvatore Coppa and Ettore Napoli},
      title = {Hardware/Software Co-Design of {ML}-{DSA} with Random Delay Insertion Against Correlation Power Analysis},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/2933},
      year = {2026},
      url = {https://eprint.iacr.org/2026/2933}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.