Benchmark source, build provenance and analysis code for "Software Stack and Workload Costs of Classical–ML-DSA Signatures: Evidence for ICP-Brasil Migration"

Abstract

Source code and build provenance for a single-host benchmark of classical signatures combined with ML-DSA-65 (liboqs 0.15.0 and OpenSSL 4.0.1), with fresh-key and retained-key regimes and a randomized standalone ML-DSA campaign comparing liboqs-python, PyCA/OpenSSL and a native OpenSSL EVP harness. Files:02_exact_source_and_build_provenance.zip: exact benchmark source used for the final run (experiment driver, pre-flight check, private liboqs and OpenSSL build scripts, Python runners, timing code, signature profiles and hybrid engine, native C EVP harness, Bouncy Castle/JMH benchmark, certificate generator, analysis script, tests), plus liboqs build configuration, commit, hashes and runtime provenance.MLDSA_supplemental_analysis_validation_v1_1.zip: post-hoc analysis code (retained/fresh comparison, workload weights, Pareto robustness, run-block trend, backend position) and the ML-DSA cross-library correctness validation.