From Pitfalls to Progress: ML-DSA Benchmarking Accepted at MAgiCS

The FORTRESS project continues to advance the state of the art in post-quantum secure systems with new research on ML-DSA benchmarking. The paper “Apples, Oranges, and Signatures: Pitfalls and Methodology in ML-DSA Benchmarking” by Sebastien Riou, Jong-Yeon Park, Liga Anwar, Axel Poschmann, and Michael Hutter has been accepted for presentation at the MAgiCS workshop (co-located with IACR Eurocrypt 2026) in Rome on May 10th, 2026.

This work addresses a critical gap in current practices, where existing benchmarking approaches often lead to misleading or non-comparable results due to the intrinsic variability of ML-DSA signature generation. By introducing a rigorous and reproducible methodology—based on standardized datasets, clearly defined metrics, and worst-case execution time (WCET) analysis—this research sets a new milestone for ML-DSA benchmarking. It enables fair comparison across implementations and provides meaningful performance insights for real-world deployment scenarios, particularly in safety-critical and resource-constrained environments.

Looking ahead, this publication represents only the first step. An extended full version will be further developed and disseminated through the FORTRESS project, where additional results, tools, and implementation insights will be shared using FORTRESS’s own methodology and framework. This will provide the community with deeper visibility into PQC performance, practical deployment considerations, and benchmarking best practices—supporting a smoother and more reliable transition toward quantum-safe secure boot and cryptographic systems.

For more information: https://magics-workshop.cs.hs-rm.de.