Why this test exists
A Cryptographic Bill of Materials is becoming a procurement artifact in India. CERT-In's Technical Guidelines on SBOM, QBOM & CBOM, AIBOM and HBOM v2.0 (09 July 2025) define what one must contain; the DST National Quantum Mission's May 2026 report recommends (advisory) that vendor CBOM submission be mandated through procurement policy from FY 2027–28; and at least one public-sector bank tender on GeM (Punjab & Sind Bank, 2026) already carries penalty-backed SBOM/CBOM clauses with quarterly refresh obligations.
That creates a familiar failure mode. When a document becomes mandatory, the market fills with tools that produce something shaped like the document. The buyer cannot tell the difference at the demo stage, because a CBOM full of confident-looking rows and a CBOM that is actually complete and verifiable look identical in a slide.
So don't evaluate the slide. Run the test.
Each question has a yes/no answer that can be demonstrated in the room. None of them requires you to understand lattice cryptography. If a vendor needs a follow-up call to answer any of the five, that is itself the answer.
Question 1 — Can I download a real CBOM right now, without a demo call?
A CBOM is machine-readable output. Any vendor that generates one has thousands of them sitting on disk, and can hand you a redacted sample in seconds. Gating the artifact behind a booked call usually means one of three things: the output is not ready, the output is thin, or the output is a screenshot.
Passing answer: a CycloneDX (or SPDX) file arrives in your inbox before the meeting ends, and it parses.
Red flag: “We'll walk you through it on the call.”
Question 2 — Is the CBOM signed? With what algorithm?
CERT-In's Table 8 (“Minimum Elements of QBOM & CBOM”) lists an Attestations element among the QBOM's elements — a digital signature, so authenticity and integrity can be established. The same reasoning applies to a CBOM, because an unsigned JSON file has no integrity property at all: any party in the chain — the vendor, an integrator, a well-meaning analyst cleaning up a spreadsheet — can change a row, and nobody downstream can tell.
Ask which algorithm. In a post-quantum context, a signature that is itself quantum-vulnerable is an awkward artifact to build a quantum-readiness programme on. ML-DSA (FIPS 204) is the standardised answer; a hybrid classical+PQC signature is a defensible one; “we don't sign” is a gap you will have to close later, at submission time, under deadline.
Passing answer: a named signature algorithm, and a public key you can fetch independently of the file.
Question 3 — Can it be verified offline, with your service unreachable?
This is the question most tools fail. A CBOM submitted to a regulator may be re-examined a year later, by a third party, during an incident, possibly after the vendor relationship has ended. Verification that requires calling the vendor's API is not verification — it is a trust assertion with extra steps.
Offline verification needs three things: a canonical serialisation (so the bytes that were signed can be reconstructed exactly — RFC 8785 JCS is the usual choice), a published or embedded public key, and a standalone verifier that runs with no network.
Passing answer: “Here is the verifier, here is the fingerprint we publish, run it on a disconnected laptop.”
Question 4 — Show me tamper detection. Change one byte.
If questions 2 and 3 were answered honestly, this one is a thirty-second demonstration: edit a single character in the CBOM, re-run the verifier, and watch it fail. Ask how a compliance pipeline consumes that result: it needs a documented, scriptable pass/fail result, not a screen someone has to read.
Passing answer: a live tamper demo and a documented, scriptable pass/fail result.
Red flag: a dashboard that shows a green tick. The green tick is rendered by the same party that produced the file.
Question 5 — What is the methodology behind your quantum timelines?
Quantum-readiness marketing is full of confident dates. Some of it is wrong in a way that matters, because it lands in board papers.
The defensible reference points are policy deadlines, not physics predictions. NIST IR 8547 (final, October 2026; all quantum-vulnerable public-key algorithms, whatever the key size) deprecates RSA-2048 and ECC P-256 after 2030 and disallows them after 2035. NSA's CNSA 2.0 sets preference dates from 2025–26 and required dates in the 2030–2033 range by category. Those are transition mandates. A vendor that renders “2030” as “the year RSA-2048 breaks” has either not read the source or has decided the distinction is inconvenient — and if they will round off that, ask what else in the report is rounded off.
Passing answer: the deadline cited to its source document, with deprecation language intact, and any break-year model shown as a model with stated assumptions.
Score sheet
| # | Question | Pass looks like | Fail looks like |
|---|---|---|---|
| 1 | Sample CBOM now, no call | A parseable CycloneDX/SPDX file | “Book a demo” |
| 2 | Signed? Which algorithm? | Named algorithm + fetchable public key | Unsigned, or “the platform secures it” |
| 3 | Offline verification | Standalone verifier, no network | Verification via vendor API only |
| 4 | Tamper demo + scriptable result | One byte changed → FAIL | A green tick in a dashboard |
| 5 | Timeline methodology | IR 8547 / CNSA 2.0 cited correctly | “RSA breaks in 2030” |
Our own answers
It would be a poor test if we did not run it on ourselves. Here is where KavachQ stands, stated so you can check it rather than believe it.
| # | KavachQ | How to check |
|---|---|---|
| 1 | Yes — CycloneDX 1.7 CBOM, on request, before any call | Ask; or run the free scan and take the output |
| 2 | Yes — ML-DSA-65 (FIPS 204), over an RFC 8785 JCS canonicalisation | The signature block embeds the public key and canonical hash; we publish the signer fingerprint separately so you can pin it |
| 3 | Yes — standalone verifier, no network, no KavachQ service required | Run it on a disconnected machine |
| 4 | Yes — the standalone verifier returns PASS/FAIL; edit one byte and it fails | Edit one byte in front of us |
| 5 | We cite IR 8547 deprecation/disallowance dates and the DST FY 2027–28 procurement date; we publish no break-year | Every date on this site carries its source |
In scope today: a CBOM (CycloneDX 1.7) signed with ML-DSA-65 and verifiable offline, plus an SBOM in CycloneDX 1.6 and SPDX 2.3 (not signed on its own); signed QBOM, AIBOM and HBOM; VEX in OpenVEX 0.2.0 and CSAF 2.0; discovery across source and configuration, cloud inventory, live TLS, keystores and certificates, and agentic machine identity.
Deliberately out of scope: CVE reachability analysis, live HSM enumeration (HSM key metadata is parsed from exports — PKCS#11, payShield), and SaaSBOM generation. Those belong to a different product category, and a tool that claims all of them well usually does none of them deeply. If your tender requires them, say so early and we will tell you plainly whether we fit.
How to use this in a procurement
Paste the five questions into the technical-evaluation annexure and require written answers. They are vendor-neutral, none of them favours a particular architecture, and each one maps to something CERT-In v2.0 already asks for. The cost of asking is nil. The cost of not asking is discovering, at submission time, that the artifact you have been paying for cannot be verified by the party you are submitting it to.
The specification our own answers are drawn from — every column, its evidence source and its owner — is published in full as the CBOM Evidence Standard. Read it before you accept our answers.
Related reading: CERT-In xBOM Guidelines v2.0, clause by clause · What a CBOM actually is · The 13-column vendor self-declaration