ML-DSA (FIPS 204) defies conventional benchmarking due to its "rejection sampling" mechanism, which causes high execution time variability. This unpredictability risks catastrophic timeouts and system failures in mission-critical, real-time environments.
Developed in collaboration with the University of Bundeswehr Munich, this white paper outlines 10 key pitfalls in existing ML-DSA benchmarking, and proposes a robust, standardized methodology.
Download the full paper to discover how to accurately evaluate worst-case execution times, compare cryptographic libraries fairly, and provision hardware for a secure quantum-safe migration
