ML-DSA timing data v1.1.0

This release adds two measured data sets and the files needed to reproduce every table and figure of the manuscript.

• Fresh-input ESP32 telemetry (measurements/telemetry/esp32_reports_fresh.csv, esp32_runs_fresh.csv): the 18 ESP32 telemetry conditions repeated with a distinct message and signing randomness for every request. All 54,000 signature digests matched host-side regeneration.
• Validation blocks (measurements/signing/rp2040/validation/): 60,000 new RP2040 signatures on the original keys and messages, captured after the calibration rule in validation/frozen_rule.json had been fixed. The session record carries that file's SHA-256.
• Analysis scripts (analysis/, run_all.sh), the NIST FIPS 204 potential-updates spreadsheet (references/), and raw capture archives with their SHA-256 (raw/).

All measurement files of version 1.0.3 are unchanged. The dataset remains under CC BY 4.0. The ZIP and its SHA-256 checksum are attached; CHECKSUMS.sha256 inside the archive identifies the individual files.

ML-DSA timing data v1.0.3

This release adds device and x86 iteration percentiles, bootstrap summaries, and the exact distribution obtained by resampling the 10 shared device keys jointly across boards, messages and boots.

All 9 device percentiles match their corrected per-set geometric references. Every pointwise 95% key-bootstrap interval contains its reference; the 99.9th-percentile intervals are [27, 27], [32, 33] and [23, 25]. The notes document the discrete estimators and conditioning.

All 41 released measurement files and every result file from version 1.0.2 remain unchanged. The README file count is corrected from 43 to 41; the earlier count included 2 local housekeeping files that were excluded from the releases.

The dataset remains under CC BY 4.0. The ZIP and its SHA-256 checksum are attached; CHECKSUMS.sha256 inside the archive identifies the individual files.

ML-DSA timing data v1.0.2

This release adds pooled iteration estimates for all 3 ML-DSA parameter sets, 95% intervals clustered on the 10 shared keys, and comparisons between the recorded rejection counts and the simplified and finite-width norm approximations. The accompanying notes state the formulas and conditioning used in these calculations.

All 43 measurement files are unchanged from version 1.0.0. The corrected deadline references introduced in version 1.0.1 are retained. The dataset remains licensed under CC BY 4.0.

The ZIP contains measurements, numerical summaries and their documentation. Its SHA-256 file identifies the archive; CHECKSUMS.sha256 inside the ZIP identifies the individual files.

ML-DSA timing data v1.0.1

This release updates the numerical results to the repetition counts proposed in NIST's July 31, 2026 potential corrections to FIPS 204. All 43 measurement files remain byte-for-byte unchanged from version 1.0.0.

The corrected common geometric reference uses a mean of 5.14, giving iteration budgets of 32, 64 and 96 for 10^-3, 10^-6 and 10^-9, and a cap reference of 821. Deadline estimates, extreme sensitivity, sample-size summaries and iteration comparisons have been recalculated. The dataset also includes per-unit and pooled ML-DSA-65 rejection totals: 100,000 signatures, 514,032 candidates and 448 hint rejections.

NIST labels these changes as proposed corrections, not an official revision of the published standard. The original 2024 repetition values remain identified in the comparison files. Updated field names are documented in DATA_DICTIONARY.md.

The release contains measurements, numerical results, field definitions, settings, source attribution and checksums under CC BY 4.0.

ML-DSA timing data v1.0.0

Version 1.0.0 of the experimental dataset by Rojin Chhetri and Babu Pillai.

The release includes 350,000 controlled ML-DSA signing records, archived signing and iteration measurements, separate stage diagnostics, and all 57 completed telemetry runs. Numerical results, experimental settings, a data dictionary and SHA-256 checksums are included.

Download mldsa-timing-data-v1.0.0.zip for the complete dataset. It contains data and documentation only. The accompanying checksum identifies the ZIP file.

Licensed under CC BY 4.0. Citation metadata are provided in CITATION.cff.