Chapter 7 — Thesis synthesis and reproducibility archive
Lesson map
Lesson 07 published. Map, example, EX039.
flowchart TD
N0["Learning outcomes and connection"]
N1["7.1 The argument map"]
N2["7.2 Structure of the thesis"]
N3["7.3 A tiered archive"]
N4["7.4 Practical activity — Lab 7: a fail-closed manife"]
N5["7.5 Dataset cards"]
N6["7.7 Implications without overreach"]
N7["Chapter synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7
N7 --> EX["Worked example"]
EX --> Q["Exercise EX039"]
Logical drawing
flowchart LR
N0["Learning outcomes and connection"]
N1["7.1 The argument map"]
N2["7.2 Structure of the thesis"]
N3["7.3 A tiered archive"]
N4["7.4 Practical activity — Lab 7: a fail-closed manife"]
N5["7.5 Dataset cards"]
N6["7.7 Implications without overreach"]
N7["Chapter synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7 --> Q["EX039"]
[Learning outcomes and connection]
|
v
[7.1 The argument map]
|
v
[7.2 Structure of the thesis]
|
v
[7.3 A tiered archive]
|
v
[7.4 Practical activity — Lab 7: a fail-closed manife]
|
v
[7.5 Dataset cards]
|
v
[7.7 Implications without overreach]
|
v
[Chapter synthesis]
|
v
[Worked example] --> [Exercise EX039]
Learning outcomes and connection
After this chapter you can (1) map a thesis argument from data to claim with explicit warrants; (2) build a tiered, fail-closed reproducibility archive; (3) write dataset cards for synthetic and restricted corpora; (4) state limits directionally — saying which way each limit would bias the result; and (5) connect the thesis to migration practice without overreach. It assembles Chapters 1–6 and prepares the defense of Chapter 8.
7.1 The argument map
Use Toulmin's structure for each major claim: data (Lab 5 paired differences), warrant (paired bootstrap estimates sampling uncertainty when items are exchangeable), backing (the method's standard use and the pilot's agreement statistics), qualifier ("where sensors observe the segment"), rebuttal (off-sensor appliances, unknown assets) and claim. Drawing the map exposes missing warrants. A common gap in PQC theses is the warrant that links inventory recall to risk reduction: it requires an argument about data lifetimes (Chapter 1's Mosca frame) and about what the organization does with findings, neither of which the inventory measures.
7.2 Structure of the thesis
A conventional order works: introduction and gap; background on quantum threat, standards and hybrid deployment; method and observer model; replication; main study; validation; critique and limits; conclusions. Put the narrowed claim in the first paragraph of the abstract. Put the architectural boundary in the method chapter, not hidden in limitations. Keep a single source of truth for every number — a results file generated by code — so that the abstract, tables and text cannot disagree.
Write the method chapter before the results chapter.
7.3 A tiered archive
Release tier 0 and tier 1 artefacts publicly; keep tier 2 with the data owner under an access agreement; never hold tier 3. The archive's manifest records each file's tier and SHA-256 hash, and the release process is fail-closed: any tier-2 file or hash mismatch blocks the release. Include the frozen configuration hash from Chapter 5 so that examiners can confirm validation used the frozen rules.
Run the release check on the exact archive that will be published, not on a working copy. Keep the manifest under version control.
7.4 Practical activity — Lab 7: a fail-closed manifest
# Lab 7 (synthetic): a fail-closed reproducibility manifest. Any tier >= 2 file in the
# public archive, or any hash mismatch, stops the release.
import hashlib, json
files = { # name: (tier, content) - contents are synthetic stand-ins
"fusion_rules.json": (0, b'{"window_days": 14, "match": "canonical-host"}'),
"corpus_synthetic.csv": (0, b"use_id,class,static,passive\nU1,code,1,0\n"),
"handshakes_pseudonymized.csv": (1, b"host,group\nh-7f3a,0x001D\n"),
"raw_capture.pcap": (2, b"\x00\x01binary"),
}
manifest = {n: {"tier": t, "sha256": hashlib.sha256(c).hexdigest()} for n, (t, c) in files.items()}
def release(manifest, files, public_max_tier=1):
problems = []
for n, m in manifest.items():
if m["tier"] > public_max_tier:
problems.append(f"{n}: tier {m['tier']} not releasable")
elif hashlib.sha256(files[n][1]).hexdigest() != m["sha256"]:
problems.append(f"{n}: hash mismatch")
return ("BLOCKED", problems) if problems else ("RELEASED", [])
print(release(manifest, files))
del manifest["raw_capture.pcap"]
print(release(manifest, files))
print(json.dumps(manifest["fusion_rules.json"])[:60] + "...")
Expected output (the hash prefix is deterministic):
('BLOCKED', ['raw_capture.pcap: tier 2 not releasable'])
('RELEASED', [])
{"tier": 0, "sha256": "db6dd8db390974cdd06e19b00d7fe35d26836...
The first call blocks the release because a tier-2 capture is in the manifest; removing it allows release. In a real archive, removing the file is not enough: check that no tier-0 or tier-1 file contains extracts from it, and record the removal in the archive's change log. The printed hash prefix lets an examiner confirm that the fusion rules in the archive match those frozen before validation. Run it again after every change to the archive.
7.5 Dataset cards
For each corpus, write a card covering: provenance (synthetic generator and seed, or owner and authorization), collection period, instruments and versions, pseudonymization method, known gaps (segments without sensors), label definitions (the amended codebook), agreement statistics, intended uses and prohibited uses (for example, no re-identification and no use as a target list). Synthetic corpora also need cards, because readers otherwise mistake generated distributions for field data.
Store each card beside its corpus and update it when errata are found. Date every card revision.
7.6 Directional limits
A limit without a direction is a ritual. State which way it pushes the estimate. Unknown assets excluded from denominators → recall overestimated (bounded by the worst-case analysis). Sensors placed where appliances concentrate → aggregate gain overestimated for organizations with sparser sensors. Adjudicators with owner documentation → ground truth may miss undocumented uses → both methods' recall overestimated, gain direction uncertain. Synthetic organizations → external validity unknown in both directions. A limits table with a "direction" column is one of the most valuable pages in the thesis.
7.7 Implications without overreach
The defensible practical implication is narrow: migration programmes that rely on static analysis alone are likely to under-find quantum-vulnerable uses in closed vendor appliances, and should combine sources where they can observe traffic and should report unknown shares. It does not follow that any organization is ready, that hybrid negotiation is downgrade-resistant, or that inventory replaces crypto-agility planning as described in NIST CSWP 39.
Write the implication as a recommendation with its condition attached, so that a practitioner reading only that paragraph still sees the scope: where sensors can observe traffic, and with unknown shares reported.
7.8 Common errors and limits
- Numbers typed by hand into the text that drift from the results file.
- Archives that silently include restricted captures.
- Limits listed without direction or magnitude.
- Implications that leap from "found more" to "safer".
Each error has a simple guard. Generate every number in the text from the results file. Run the fail-closed manifest check on the release archive. Add a direction and, where possible, a magnitude to each limit. Read the implications paragraph aloud and stop at any verb that implies safety rather than discovery.
Exercises — chapter 7
- EX039 — Toulmin map. Map the H4 claim with all six elements. Rubric: 3 = warrant and rebuttal specific.
- EX040 — Missing warrant. Write the argument linking recall to risk reduction and list what evidence it would need. Rubric: 3 = involves data lifetime and remediation.
- EX041 — Manifest. Extend Lab 7 to fail when a file listed in the manifest is missing. Rubric: 3 = code works and fails closed.
- EX042 — Dataset card. Write the card for the Marrow corpus. Rubric: 3 = includes prohibited uses and known gaps.
- EX043 — Directional limits. Add three limits with direction and magnitude to the table in 7.6. Rubric: 3 = directions argued, magnitudes bounded where possible.
- EX044 — Implication paragraph. Write the practitioner implication for a migration lead in 150 words. Rubric: 3 = useful and within the evidence.
Chapter synthesis
A thesis is an argument whose warrants are explicit, whose numbers have one source, whose archive cannot leak restricted data, and whose limits point in stated directions. Its implications are as narrow as its claims.
Continue studying — chapter 07 of 08
← Chapter 06 (draft) · Course chapter index · Chapter 08 (draft) →