Your RSA-2048 keys break in 2030. Find every one of them before attackers do.
Compliance›CERT-In SBOM/CBOM Guidelines
India · CERT-In v2.0 (09 Jul 2025)

CERT-In SBOM, CBOM, QBOM, AIBOM & HBOM

Technical Guidelines v2.0 — what each bill of materials must contain and how to produce it

Published Last reviewed

On 09 July 2025 CERT-In, the national CERT under MeitY, published Version 2.0 of its Technical Guidelines on SBOM, QBOM & CBOM, AIBOM and HBOM. It is the canonical Indian reference for software-supply-chain transparency: it defines the minimum elements of a Software Bill of Materials (Table 5), aligns SBOM generation to the SDLC (§3.2), mandates VEX and CSAF for vulnerability tracking (§6), and extends the model to cryptography (CBOM/QBOM, §8), AI systems (AIBOM, §9) and hardware (HBOM, §10).

The guidance is recommended for public-sector, government, essential-services and software-export organisations, and it makes SBOM a standard practice in software procurement. Sectoral regulators including SEBI and RBI point to it as the field-level reference for their own SBOM obligations.

At a glance
09 Jul 2025Version 2.0 published (v1.0 Oct 2024)
5 BOM typesSBOM, CBOM, QBOM, AIBOM, HBOM
6 SBOM classesDesign, Source, Build, Analyzed, Deployed, Runtime
21 / 11 / 19 / 18Minimum elements: SBOM, QBOM, AIBOM, HBOM
CERT-In Technical Guidelines on SBOM, QBOM & CBOM, AIBOM and HBOM — Version 2.0
Why this matters

India's reference standard for software-supply-chain transparency

Software supply-chain attacks reach organisations through trusted dependencies: malicious code, vulnerabilities in outdated components, and breaches via compromised suppliers. CERT-In v2.0 answers this by making a machine-readable inventory of what ships in software a baseline expectation, particularly for public-sector, government, essential-services and software-export entities.

The document is broader than a classic SBOM. It treats the bill of materials as a family: SBOM for software components, CBOM for cryptographic assets, QBOM for quantum-readiness and post-quantum migration (§8), AIBOM for AI models, frameworks and training datasets (§9), and HBOM for physical hardware (§10). Vulnerability state is communicated through VEX and CSAF rather than raw scan dumps (§6).

Each type carries its own minimum-elements table, and the lists do not overlap. An SBOM needs the 21 data fields of Table 5; a QBOM needs the 11 elements of Table 8 plus per-asset detail from Table 9; an AIBOM needs the 19 elements of Table 10; an HBOM needs the elements of Table 11, recursed into sub-components. Treating Table 5 as the universal checklist is the common error, and it leaves a supplier deliverable short in exactly the places the newer sections were written to cover.

Although the guidelines are advisory rather than statute, they have become the field-level reference that SEBI's CSCRF SBOM mandate and RBI's supply-chain expectations lean on. Adopting them now positions an organisation ahead of the sectoral rules that cite them.

How O3 helps

Find it before attackers do. Catch it if they try. Fix it fast.

Across these requirements, O3 runs one continuous loop on your own software so issues surface, get caught, and get fixed before they become an incident.

1 · Find it first

Test like an AI-assisted attacker

An agentic pentest probes your app across 50+ vulnerability classes the way a real attacker chains them, so exploitable flaws surface in your build before someone outside finds them.

2 · Catch it live

Detect and block exploitation at runtime

A kernel-level eBPF agent watches process trees and syscalls, recognises an attack sequence as it unfolds, and restricts the offending process on the spot rather than taking down the host.

3 · Fix it fast

Auto-patch what is reachable

Reachability ranks what genuinely matters, then autofix opens a pull request with the change, so remediation starts inside tight fix windows instead of sitting in a queue.

Risk-based prioritisation

From alert flood to a short list

KEV + EPSS + reachability turn a raw scan into the few findings that actually need action now.

Raw findings in dependency tree~12,400
Vulnerable & version-matched~4,700
Reachable in your code~1,360
KEV / high-EPSS · act now~370

Illustrative funnel. Reachability removes the bulk of unexploitable noise before triage.

SBOM minimum elements

How to build an SBOM that meets CERT-In's minimum elements

Table 5 of v2.0 defines three minimum elements: Data Fields, Automation Support, and Practices and Processes. The Data Fields element alone lists 21 attributes to record for every component. A compliant SBOM is machine-readable and complete down to transitive dependencies, not a hand-maintained spreadsheet.

Capture, per component: name, version, a unique identifier (PURL or CPE), supplier or author, and the dependency relationship including top-level and transitive dependencies (Table 5).
Record a cryptographic hash and a timestamp for each component so the SBOM is verifiable and point-in-time.
Document license information for every component and flag any 'known unknowns' where completeness cannot be established.
Emit in a standard machine-readable format such as CycloneDX or SPDX, and regenerate on every build rather than editing by hand.
Where O3 helps

O3 generates CycloneDX SBOMs with component identifiers, transitive dependency graphs, licenses and cryptographic hashes, and enriches them with deps.dev metadata so the Table 5 fields are populated automatically.

SDLC SBOM classes

How to generate the right SBOM at each SDLC stage

Section 3.2 defines six SBOM classes tied to the software lifecycle. Each answers a different assurance question, and producing the later classes is where most teams have gaps.

CERT-In v2.0 §3.2SCA tooling
Map your pipeline to the six classes: Design, Source, Build, Analyzed, Deployed and Runtime SBOMs (§3.2).
Generate Source and Build SBOMs inside CI using Software Composition Analysis (SCA) tooling, which the guidelines explicitly endorse for generation and vulnerability detection.
Produce an Analyzed SBOM by running SCA and vulnerability analysis against the build output, and a Deployed SBOM reflecting what actually runs in production.
Capture a Runtime SBOM from the live environment so loaded components, not just declared ones, are inventoried.
Where O3 helps

O3's SCA produces Source, Build and Analyzed SBOMs in CI, and its eBPF runtime agent observes loaded components and processes in production to support a Runtime SBOM that reflects what is actually executing.

Vulnerability tracking

How to track and communicate vulnerabilities with VEX and CSAF

Section 6 requires that vulnerability state be expressed through VEX (Vulnerability Exploitability eXchange) and CSAF (Common Security Advisory Framework), using the Log4j incident as a worked example.

CERT-In v2.0 §6VEXCSAF
For each known vulnerability affecting an SBOM component, publish a VEX statement declaring whether the product is affected, not affected, fixed or under investigation (§6).
Issue machine-readable advisories in CSAF format so downstream consumers can ingest your security posture automatically.
Prioritise exploitable issues: combine reachability with EPSS exploit-probability and CISA KEV known-exploited status before declaring something affected.
Aim to publish VEX promptly after disclosure; the v2.0 Log4j example illustrates advisory publication within roughly a week.
Where O3 helps

O3's function-level reachability determines whether a vulnerable component is actually reachable, and EPSS plus KEV prioritisation drives the affected vs not-affected judgement that a VEX statement records.

CBOM, QBOM & PQC

How to inventory cryptography and prepare for post-quantum migration

Section 8 extends the BOM model to cryptography, and it splits the requirement in two. Table 8 sets eleven minimum elements for a QBOM, describing a quantum device or system. Table 9 sets the elements for each cryptographic asset inside it, and those differ by asset type — an algorithm is documented differently from a key, a protocol or a certificate.

Start with the device or system. Table 8 asks for eleven elements: model name, version, vendor and origin, licence, cryptographic asset, communication protocol, hardware, software dependencies, environmental impact, vulnerabilities and attestations.
Algorithms need name, asset type, primitive, mode, crypto functions, classical security level and OID. CERT-In's worked example is AES-128-GCM, OID 2.16.840.1.101.3.4.1.6 — the OID is what makes the entry matchable across tools.
Keys are a separate shape: id, state, size, creation date, activation date. State is the field auditors ask about, because a revoked key still sitting in a config is a finding and an active one is not.
Protocols carry version and cipher suites, so TLS 1.2 and TLS 1.3 are two entries rather than one "TLS" row. Collapsing them hides exactly the downgrade you are trying to find.
Certificates need subject and issuer names, the not-valid-before and not-valid-after window, signature algorithm and subject public key references, format and extension.
With the inventory in place, flag the quantum-vulnerable algorithms — RSA and ECC — and sequence a PQC transition, tracking it through the QBOM.
Re-scan on a schedule. Certificates expire and libraries rotate, so a CBOM that was accurate at procurement is stale a quarter later.
Where O3 helps

O3 generates a CycloneDX CBOM of cryptographic assets and a QBOM with quantum-risk scoring, surfacing quantum-vulnerable algorithms to drive a PQC migration roadmap.

AIBOM

How to produce an AIBOM for AI systems

Section 9 introduces the AI Bill of Materials. Table 10 sets nineteen minimum elements, and they go well beyond a component list: the AIBOM is expected to document what the model is for, what it must not be used for, and how it performs.

The model first: name, version, type (text-generation, image-classifier and so on) and the developer or organisation behind it.
Licensing covers the model and every component used, and sits alongside software dependencies — third-party libraries, frameworks, OS requirements, proprietary software.
Name the ML models and algorithms in the system, then the metrics used to judge them: accuracy, precision, recall, F1.
Data source and data sets are two different fields. Source is the origin — proprietary, public, real-time, synthetic. Data sets carries names, versions, formats and limitations, and that is the entry a privacy or licensing review actually reads.
Hardware means what the model needs to run: GPUs, TPUs, storage, memory. Security requirements sit next to it — the encryption and access control protecting model data and user information.
Input and output types, intended usage, and out-of-scope usage. That last one is unusual and worth reading twice: CERT-In wants the uses a model is not built for written down, so misuse is a documented deviation rather than a judgement call after the fact.
Finish with environmental impact, known vulnerabilities carrying severity and CVE references, and a digital attestation over the AIBOM.
Where O3 helps

O3 generates a CycloneDX AIBOM that inventories AI models, frameworks and datasets, giving teams the model and data component list the AIBOM section calls for.

HBOM & provenance

How to cover hardware and component provenance

Section 10 defines the Hardware Bill of Materials. Table 11 records the product and each component beneath it, and it is explicitly recursive: where a component contains further hardware parts, those are listed with the same attributes.

The product level is straightforward: name, version, details, warranty or AMC terms.
Provenance is where it gets specific: manufacturer name, manufacturer location, manufacturing date, then supplier information and supplier location. Both appear twice in Table 11, once for the product and once per component, because the company that made a part is rarely the one that sold it to you.
Model number identifies the design; serial number identifies the individual unit. Auditors use them differently, so both are required.
Each component needs its technical specification (voltage, frequency, capacity) and, for integrated circuits, the technology node: 7nm or 14nm, which pins the generation.
Compliance declarations (RoHS, CE), power supply requirements, and licence information for any firmware or embedded software inside the component.
Test results record operational status from QA. Then recurse: where a component contains further hardware parts, those get the same attributes again.
When a component vulnerability surfaces, §10.4.1.5 asks for a VEX or equivalent hardware vulnerability statement — not affected, affected, fixed, under investigation — followed by a CSAF advisory carrying severity, mitigations and impact.
Where O3 helps

O3 reads SLSA and cosign provenance and verifies signed attestations during ingestion, supporting the provenance checks the guidelines call for; it does not generate provenance itself.

Procurement & maintenance

How to make SBOM a standard part of procurement

The guidelines make SBOM a standard practice in software procurement and expect bills of materials to be kept current rather than produced once.

CERT-In v2.0 procurement guidance§3
Require an SBOM as a contractual deliverable for new software acquisitions, in a standard machine-readable format.
Validate supplier-provided SBOMs against the Table 5 minimum elements before acceptance.
Regenerate and re-verify the SBOM on every software change, upgrade or patch so it never goes stale.
Run continuous SCA and malicious-dependency checks against the maintained SBOM to catch newly disclosed or compromised components.
Where O3 helps

O3's SCA and malicious-dependency detection (typosquats, malicious postinstall scripts, compromised maintainers) continuously validate the components recorded in an SBOM as new risks emerge.

The whole framework

Every BOM type and requirement in CERT-In Technical Guidelines v2.0

All 40 requirement areas, each with the reference and a vendor-neutral note on how teams meet it.

12SBOM
Minimum elements and SDLC classes
2Vulnerability
VEX and CSAF tracking
3Cryptography
CBOM, QBOM and PQC
4AI & Hardware
AIBOM and HBOM
2Supply chain
Provenance and procurement
Every BOM type and requirement in CERT-In Technical Guidelines v2.0 — 40 requirement areas, each with its reference and how teams meet it.
Requirement areaRef
Component NameTable 5
Component VersionTable 5
Component DescriptionTable 5
Component SupplierTable 5
Component LicenseTable 5
Component OriginTable 5
Component DependenciesTable 5
VulnerabilitiesTable 5
Patch StatusTable 5
Release DateTable 5
End-of-Life (EOL) DateTable 5
CriticalityTable 5
Usage RestrictionsTable 5
Checksums or HashesTable 5
Comments or NotesTable 5
Author of SBOM DataTable 5
TimestampTable 5
Executable PropertyTable 5
Archive PropertyTable 5
Structured PropertyTable 5
Unique IdentifierTable 5
Automation SupportTable 5
Known UnknownsTable 5
SDLC SBOM classes§3.2
SCA tooling§3
Machine-readable format§3
VEX documents§6
CSAF advisories§6
QBOM minimum elementsTable 8
Cryptographic asset elementsTable 9
Crypto-agility§8
PQC transition plan§8
AIBOM minimum elementsTable 10
Intended / out-of-scope usageTable 10
Model performance metricsTable 10
HBOM minimum elementsTable 11
Technology nodeTable 11
Hardware VEX + CSAF§10.4.1.5
Provenance monitoringThreats
Procurement practiceProcurement

Generate every BOM type CERT-In v2.0 asks for

O3 produces CycloneDX SBOM, CBOM, QBOM, AIBOM and HBOM, runs SCA with malicious-dependency detection, ranks exploitability with function-level reachability plus EPSS and KEV, and observes runtime components via its eBPF agent. See how it maps to CERT-In's minimum elements, SDLC SBOM classes and VEX/CSAF tracking.

Read the v2.0 guidelines on cert-in.org.in
FAQ

Common
questions.

Everything teams ask before rolling this out. Still stuck? Reach our team.

  • They are Technical Guidelines published by CERT-In under MeitY on 09 July 2025, defining how to build a Software Bill of Materials and four related bills of materials: CBOM and QBOM for cryptography (§8), AIBOM for AI systems (§9) and HBOM for hardware (§10). Table 5 lists the SBOM minimum elements.
  • The CERT-In guidelines are advisory rather than statutory, but they make SBOM a standard practice in software procurement and are recommended for public-sector, government, essential-services and software-export entities. Sectoral regulators such as SEBI's CSCRF impose explicit SBOM mandates that reference these field-level requirements.
  • Table 5 of CERT-In's v2.0 guidelines defines three minimum elements: Data Fields, Automation Support and Practices and Processes. The Data Fields element lists 21 attributes to record per component: name, version, description, supplier, license, origin, dependencies, vulnerabilities, patch status, release date, end-of-life date, criticality, usage restrictions, checksums or hashes, comments or notes, author of the SBOM data, timestamp, executable property, archive property, structured property and unique identifier. Automation Support requires a machine-readable format (SPDX or CycloneDX), and Practices and Processes covers frequency, depth, known unknowns, distribution and delivery, access control and accommodation of mistakes.
  • Section 3.2 aligns SBOMs to the SDLC: Design, Source, Build, Analyzed, Deployed and Runtime SBOMs. Each captures a different stage, with Build and Analyzed SBOMs generated in CI using SCA tooling and a Runtime SBOM reflecting components actually loaded in production.
  • Section 6 requires vulnerability state to be communicated through VEX (Vulnerability Exploitability eXchange) statements declaring whether a product is affected, and CSAF (Common Security Advisory Framework) advisories in machine-readable form. The Log4j example shows VEX publication within roughly a week of disclosure.
  • Section 8 defines a CBOM as an inventory of cryptographic assets such as algorithms, keys and protocols, and a QBOM addressing quantum readiness. Organisations are expected to assess crypto-agility, identify quantum-vulnerable algorithms, and plan a Post-Quantum Cryptography migration.
  • Section 9 requires an AI Bill of Materials covering the components used to build, train and deploy AI models: hardware such as GPUs and servers, software including model name and frameworks, and the datasets or training data used. It makes AI supply-chain composition transparent.
  • O3 generates CycloneDX SBOM, CBOM, QBOM, AIBOM and HBOM, runs SCA with malicious-dependency detection, ranks exploitability using function-level reachability plus EPSS and KEV to inform VEX, observes runtime components via its eBPF agent, and reads SLSA and cosign provenance. Governance and policy steps remain the organisation's responsibility.

See O3 Security in Action

See why security and engineering leaders trust O3
to secure their entire software supply chain.