Dataset Card — PQC-MFB

Summary

322 executed cases describing how post-quantum migrations of authenticated key-exchange protocols fail. Each case pairs an unrepaired ("naive") design with a failure family and the protocol invariant that breaks. 312 of the 322 are failures; the remaining 10 pass even unrepaired and serve as regression controls.

• Cases: 322 (312 failures, 10 controls)
• Failure families: 39
• Unrepaired designs: 10
• Format: JSON Lines, one case per line
• Licence: CC-BY-4.0 (data) / Apache-2.0 (code)

Fields

Field	Type	Description
case_id	string	<design>::<family>, unique
design	string	which unrepaired design was executed
family	string	failure family (one of 39)
invariant	string	the protocol invariant under test
is_failure	bool	whether the unrepaired design fails this case
naive_detail	string	what the unrepaired design actually did
naive_held	bool	whether the invariant held unrepaired
prior_art_analogue	string	a published failure of similar shape, where one exists

Deliberately withheld

repair_mechanism, repaired_detail and repaired_held are removed from the public release. Together they form the failure-to-mechanism incidence matrix, which is the input to a minimum-cover computation over the repair set and belongs to the closed core.

Scoring does not need them: the benchmark asks whether your implementation holds the invariant, not which mechanism you used. The projection step is src/pqc_mfb/build_dataset.py, and a test asserts the withheld fields never appear in the published file.

How it was produced

Cases were executed against a modeled protocol state machine — not captured from network traffic or from shipping products. For each (design, family) pair the harness runs the adversary for that family against the unrepaired design and records whether the named invariant held.

The public file is derived from the internal manifest by a single reproducible projection; regenerate it with:

python -m pqc_mfb.build_dataset <internal_manifest.json> src/pqc_mfb/data/pqc_mfb.jsonl

Intended use

• Scoring a post-quantum AKE implementation for migration-failure coverage.
• Teaching: a concrete catalogue of what breaks when key material outgrows a frame.
• Research on protocol state-machine failure taxonomies.

Out of scope

• Not a vulnerability disclosure. No case is an assertion about a named product.
• Not a security certification. A perfect score means 312 modeled cases handled.
• Not byte-exact CVE reproduction. prior_art_analogue names a failure of similar shape. It is a pointer for the reader, not a claim of equivalence.
• Not a fuzzing corpus. There are no packet captures or binary inputs here.

Limitations and biases

• Cases derive from one modeled state machine, so they inherit its modelling choices. A failure mode that model cannot express is not represented.
• Coverage across families is uneven — some families have one case, others many. Overall coverage therefore weights families unequally, which is why the scorer also reports per-family results and names zero-coverage families explicitly.
• The 10 unrepaired designs are constructed baselines, not third-party implementations.
• Wireless-derived. Families transfer to other transports, but the designs are wireless.

Citation

@misc{pqcmfb2026,
  title  = {PQC-MFB: A Post-Quantum Migration Failure Benchmark},
  author = {{PQC Migration Safety Lab}},
  year   = {2026},
  note   = {322 cases, 39 failure families, 10 unrepaired designs}
}

Contact

Issues and submissions via the repository. For the closed core, open a GitHub Discussion or an issue.

---

The PQC migration toolkit

Nine free tools for teams moving authenticated key exchange to post-quantum. They find and measure; they do not repair.

Tool	What it does	Where
pqc-sizes	Sizes, fragment counts, and the two-sided reassembly window	PyPI
pqc-sizes-js	The same arithmetic for Node and the browser	npm
pqc-guard-action	Fail the build when the window is empty	GitHub Action
pqc-dos-embedded	169 lines of C: the failure on a real 64 KB device	source
farkas-check	Re-verify the bound on-device, no SMT solver	source
pqc-migration-mcp	Six MCP tools for AI agents	PyPI
pqc-mfb	322 cases · 39 failure families · scorer	PyPI
pqc-mfb (data) ← you are here	The benchmark as a dataset	HF
pqc-formal-corpus	122 named formal results, 6 provers	HF
pqc-explorer	Try it in your browser, no install	HF Space

Start here: pqc-sizes tells you in five seconds whether your credential fragments and whether a safe cap exists. pqc-explorer does the same in a browser.

The closed core

Closing the 39 failure families — downgrade binding, retransmission-safe installation, fragmentation transcripts, roaming forward secrecy, multi-link key separation, admission control, group-key binding — is a separate proprietary codebase. Relevant subject matter is covered by a filed provisional patent application.

That split is measured, not asserted: under a replicate noise control only 4 of 32 repair mechanisms are externally distinguishable, so publishing these detectors does not disclose the repairs.

For commercial licensing, open a GitHub Discussion or an issue on any of these repos.