CBOM vs SBOM, in one paragraph
An SBOM lists software components: names, versions, suppliers, licenses, dependency relationships. It will tell you that OpenSSL 3.0.11 is present. It will not tell you that your payments service negotiates RSA-2048 for a certificate protecting Aadhaar-linked records, that an internal CA still signs with SHA-1, or that a batch job hard-codes 3DES. That cryptographic layer is what a CBOM inventories — and it is the only layer that matters when the question is “what breaks when RSA does?”
A list is not enough — the graph is the product
The naive CBOM is a spreadsheet: these systems use RSA-2048, those use ECDH P-256. Necessary, but it cannot answer the one question a migration programme exists to answer: if this algorithm falls, what is the blast radius?
Answering that needs typed relationships — an application uses a certificate, a certificate signs with an algorithm, an algorithm protects a class of data. With those edges you can trace from one weak algorithm to the regulated data it exposes and the internet-facing service that makes it reachable. Without them you have an inventory that tells you what you own and nothing about what you risk. (Longer treatment: The CBOM in the Knowledge Bank.)
What goes in one
CERT-In's Technical Guidelines v2.0 set out the minimum elements by asset type. Summarised:
| Asset type | Minimum elements |
|---|---|
| Algorithm | Name, asset type, primitive, mode, cryptographic functions, classical security level, OID |
| Key | Name, asset type, id, state, size, creation date, activation date |
| Protocol | Name, asset type, version, cipher suites, OID |
| Certificate | Subject, issuer, validity window, signature algorithm reference, subject public key reference, format, extensions |
Table 9 itself has no signature element. The “Attestations” row — a digital signature to ensure authenticity and integrity — sits in Table 8, among the elements of the QBOM. A signature is still what turns a CBOM into evidence a third party can check; more on that below.
The standard container is CycloneDX with its cryptographic-asset model (1.7 is current; 1.6 files are still widely accepted). SPDX is also accepted (§8.4.1.5). The format matters less than the fields being genuinely populated from observation.
Where the assets actually hide
Most teams underestimate the number of surfaces. A realistic CBOM has to reach at least these:
| Surface | What lives there |
|---|---|
| Source code | Crypto API calls, hard-coded algorithm names, key sizes, modes |
| Configuration | TLS cipher suites, JVM security settings, nginx/Apache directives, IaC |
| Certificates & keystores | PEM/DER chains, JKS and PKCS#12 stores, signature algorithms, expiry |
| Live endpoints | What is actually negotiated on the wire — often not what config claims |
| Cloud services | KMS key specs, ACM certificates, load-balancer TLS policies, Key Vault contents |
| Machine identity | OIDC issuers, JWKS keys, service-account and workload tokens signing with RS256 |
That last row is newer and largely unaddressed. As estates fill with agents and service-to-service automation, the crypto that matters most is often the signing key behind an identity provider — one anchor whose compromise forges every identity beneath it.
Why two tools disagree by orders of magnitude
Coverage is not a given, and it is the single most useful thing to test before you buy anything. A scanner that recognises a handful of language-specific crypto APIs will return near-nothing on a codebase whose cryptography lives in configuration, shell tooling, embedded certificates and keystores — and it will return that near-nothing confidently, as a clean report.
The lesson: “we generate CBOMs” is not a capability statement. Ask for finding counts on your code, from two tools, and compare.
What makes a CBOM trustworthy
- Coverage you can test. Run it on a repository whose cryptography you already know. Count what it misses.
- Honesty about the unobserved. Fields that could not be resolved should be absent, not guessed. An invented OID is worse than a missing one, because a signature would then attest to a fiction.
- A signature a third party can verify offline. CERT-In lists a digital signature among the minimum elements of the QBOM and the AIBOM for a reason, and the same logic applies to a CBOM. An unsigned CBOM cannot survive the journey through an integrator, a compliance team and a regulator with its integrity intact.
The five questions that operationalise this are on the Declaration Test page.
The clocks
| Date | What happens | Source |
|---|---|---|
| 09 Jul 2025 | CERT-In xBOM guidelines v2.0 define CBOM contents | CERT-In / MeitY |
| 2026 | At least one public-sector bank tender on GeM (Punjab & Sind Bank, 2026) requires penalty-backed SBOM/CBOM submissions | GeM bid GEM/2026/B/7366718 |
| FY 2027–28 | DST roadmap recommends mandating vendor CBOM submission through procurement policy (advisory) | DST National Quantum Mission report, May 2026 |
| After 2030 | RSA-2048 / ECC P-256 deprecated | NIST IR 8547 (final, October 2026; all quantum-vulnerable public-key algorithms, whatever the key size) |
| After 2035 | Disallowed | NIST IR 8547 |
Note what those last two rows say and do not say. They are transition deadlines set by a standards body. They are not a prediction that a quantum computer factors RSA-2048 in 2030, and material that presents them that way should be read with care.
How KavachQ produces one
Several collection paths feed one graph: live TLS and SSH probes, and your external attack surface once you prove you own the domain; passive capture, including OT protocols; an in-network collector for systems the internet cannot see; code, configuration, keystores, certificates and Git organisations; exports from cloud, Kubernetes, databases, HSMs, mainframe, payments, VPN and mobile; and imports of other tools' CBOMs and SBOMs. The full list is on KavachQ Discover. Output is a CycloneDX 1.7 CBOM, signed with ML-DSA-65 (FIPS 204) over an RFC 8785 canonicalisation, verifiable offline by a standalone verifier.
In scope today: the collection paths above, output as a CBOM (CycloneDX 1.7) signed with ML-DSA-65, plus an SBOM in CycloneDX 1.6 and SPDX 2.3 (not signed on its own); signed QBOM, AIBOM and HBOM; and VEX in OpenVEX 0.2.0 and CSAF 2.0. Deliberately out of scope: live HSM enumeration — HSM key metadata is parsed from exports (PKCS#11, payShield); no live HSM enumeration. Listed here rather than discovered by you in week three of a pilot.
Related: CERT-In xBOM Guidelines v2.0 · The CBOM Evidence Standard · The 13-column vendor self-declaration · Cryptographic discovery