Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 21 of the 22 references it cites were found in a registry or an index; 11 of the 54 sources its claims cite were checked for support.

AI Supply Chain Security: MBOM-PQC Provenance, PQC Attestation, and a Maturity Model for Quantum-Resistant Assurance

Open access
Campbell, Robert · 2026 · Systems

Abstract

A formal Model Bill of Materials with PQC-safe extensions, a unified signing and attestation pipeline, and a maturity model are proposed for AI supply chain integrity.

Overview

CoChat Summary

TL;DR

The paper synthesizes 54 policy, standards, academic, and incident sources to argue that AI supply chains lack verifiable provenance and that classical signatures used for model lineage and attestation will fail against "harvest-now, forge-later" quantum attacks. It proposes three prescriptive artifacts: the MBOM-PQC provenance schema, a PQC-safe signing and attestation pipeline built on ML-DSA/SLH-DSA and hybrid signatures, and the five-level SCAMM maturity model.

Key findings

  • Existing frameworks (NIST AI RMF, NIST SSDF, SP 800-204D, DoD CDAO RAI Toolkit, SPDX/CycloneDX SBOM standards, and PQC transition guidance such as SP 800-208 and CNSA 2.0) address adjacent concerns but none defines a verifiable AI model provenance structure or PQC-safe signing/attestation pipeline.
  • The evidence synthesis screened 142 records, deduplicated to 123, screened to 76 full-text reviews, and included 54 sources: 31 policy/standards documents, 15 peer-reviewed academic papers, and 8 documented supply chain incidents (28 of the 54 are cited directly in the manuscript body).
  • The MBOM-PQC schema is organized into seven components (model metadata; pre-training dataset lineage; pre-trained model dependencies; fine-tuning artifacts; training environment and pipeline; deployment packaging; cryptographic integrity fields), with three PQC extensions: hybrid signature bundles, PQC-safe certificate chains, and long-term integrity anchors.
  • The signing pipeline defines five stages (Ingestion, Verification, Signing, Attestation, Deployment) and recommends ML-DSA-65 as the default parameter set, citing a signature size of 3,309 bytes (~3.3 KB) versus 64 bytes for Ed25519, with SLH-DSA signatures ranging roughly 8–50 KB depending on parameter set; SCAMM defines five cumulative levels from Ad Hoc (L1) to Continuously Attested/Zero Trust-aligned (L5).

Methods

  • Dual-method design: systematic evidence synthesis (PRISMA 2020 adapted for policy, standards, and security-incident sources) combined with prescriptive architectural development; the framework artifacts are design-science outputs intended for future operational validation rather than empirical evaluation.
  • Evidence collected January–December 2025 across five source classes (standards bodies, government policy, incident records, academic literature, industry reports) via searches on IEEE Xplore, ACM Digital Library, IACR ePrint, NIST CSRC, NSA.gov, CISA.gov, and curated incident repositories; inclusion required 2020–2026 publication, relevance to AI supply chain risk/cryptographic integrity/PQC transition, normative or technical content, and extractable provenance/signing/attestation content, with four exclusion codes (E1–E4).
  • Sources were coded with a nine-field extraction template and assigned confidence tiers (A/B/C); extracted requirements were mapped to schema fields, pipeline components, and SCAMM indicators via traceability matrices (Tables 1 and 2), with full materials in Supplementary Materials.

Limitations

  • Stated: PQC algorithms and certificate formats continue to evolve, so specific parameter-set recommendations may need revision as NIST/IETF guidance matures; the framework lacks empirical validation through large-scale operational deployments; it depends on upstream providers supplying sufficient metadata and signatures, which varies widely; and continuous learning systems create high-velocity signing/attestation overhead that may require batched, delta-signing, or checkpoint-based adaptations.
  • Evident from the text: the version provided is a Preprints.org preprint explicitly marked "Article Not peer-reviewed version," so it has not undergone peer review; the SCAMM dashboard (Figure 10) is described as illustrative/representative rather than derived from a real organizational assessment; and the SHA-256 hash shown in Figure 6 is noted as illustrative only.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/systems14050593

Cites

21 of 22 references verified1 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

2 Valid · 21 Not valid · 31 Not checked · 81 Own results and interpretations

36 more claims

Create a free account to see all 135 claims and to verify the 40 references still open.

See all claims — free account
  • It may state two separate claims as one. 51
  • Probably not a substantial claim. 72

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 16 references

Research works printed but not linked: 2

  1. NIST (2024). Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. · DOI · Paper page
  2. Li, Y., Lyu, X., Koren, N., Lyu, L., Li, B., and Ma, X. (2021). Anti-Backdoor Learning: Training Clean Models on Poisoned Data. Proceedings of the 35th Conference on Neural Information Processing Systems (NeurIPS 2021), Curran Associates.

Metadata

Authors
Campbell, R.
Affiliations
Independent Researcher, Upper Marlboro, MD 20774, USA
Corresponding author
Campbell, Robert
Venue
Systems
Publisher
MDPI AG
Published
2026-05-21
Volume
14
Issue
5
Pages
593
ISSN
2079-8954 (ISSN-L)
OpenAlex
W7162027274
Cited by
3
References
16
Citations by year
2026: 3
FWCI
10.25
Type
journal article
Language
en
Open access
Locations
  • Systems (publishedVersion, cc-by)
  • Preprints.org (acceptedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
  • Physical Sciences › Computer Science › Computer Networks and Communications › Cryptography and Data Security
  • Physical Sciences › Computer Science › Information Systems › Blockchain Technology Applications and Security
Keywords
model provenance, post-quantum cryptography, digital signatures, ML-DSA, model poisoning, AI governance, root of trust

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: AI Supply Chain Security: MBOM-PQC Provenance, PQC Attestation, and a Maturity Model for Quantum-Resistant Assurance. https://app.cochat.ai/papers/10.3390/systems14050593
The paper
Campbell, R. (2026). AI Supply Chain Security: MBOM-PQC Provenance, PQC Attestation, and a Maturity Model for Quantum-Resistant Assurance. Systems. https://doi.org/10.3390/systems14050593

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.