Certificates are the most widely deployed machine identity in most organisations, yet few can say how many they have, who owns them, or whether they could replace all of them in a week. This maturity model gives a five-level scale, from Ad hoc to Optimised, across six capability areas: inventory and discovery, ownership and policy, issuance and automation, short-lived certificates and renewal, key protection and revocation, and crypto agility. Each level is defined by evidence an assessor can check. It is written for PKI owners, platform and security architects and CISOs who need a baseline, a defensible target and a roadmap, particularly now that the maximum lifetime of public TLS certificates is falling to 47 days. For the discipline itself see the Machine Identity, PKI and Certificate Lifecycle Guide; this page covers staged maturity, not the basics.
Key takeaways
- The calendar is setting the target: under the CA/Browser Forum Baseline Requirements, publicly trusted TLS certificates issued from 15 March 2026 may last at most 200 days, from 15 March 2027 at most 100 days and from 15 March 2029 at most 47 days, so manual renewal stops being viable for public certificates during this period.
- Score the weakest area, not the average: perfect automation for web servers does not help if an unknown internal CA signs certificates for production databases with no owner, so the overall rating is the lowest area score.
- Evidence decides the level: a discovery report, a renewal success rate or a timed mass-replacement exercise counts; a policy document or a tool purchase on its own does not.
- Level 3 is the minimum for 2027: Defined means a complete inventory with owners and automated renewal for public certificates, which is what 100-day and then 47-day lifetimes require.
- Agility is the top of the scale, not an extra: the ability to replace a CA, an algorithm or every certificate quickly is what protects against CA distrust events, key compromise and the move to post-quantum algorithms.
The five levels
| Level | Name | Description |
|---|---|---|
| 1 | Ad hoc | Certificates are requested by whoever needs one, from whichever CA is convenient, and tracked in spreadsheets or calendars if at all. Expiry is discovered by outage. Private keys sit in files on servers and in repositories. Nobody can produce a complete list. |
| 2 | Developing | A central team runs some discovery and an approved CA list exists. Some renewals are automated, mainly public web certificates. Ownership is recorded for part of the estate. Internal CAs and non-TLS certificates (code signing, device, client authentication) are largely outside the programme. |
| 3 | Defined | A certificate policy with named owners applies across public and private PKI. Continuous discovery feeds a single inventory. All public TLS certificates renew automatically, usually via ACME. Keys for CAs and signing certificates are in hardware. Revocation and replacement procedures are documented and tested. |
| 4 | Managed | The programme is measured: inventory coverage, renewal success, time to replace and policy exceptions are reported against targets. Internal certificates are short-lived where platforms allow. Renewal windows follow CA signals such as ACME Renewal Information. Mass replacement has been exercised against a deadline. |
| 5 | Optimised | Certificates are issued on demand for workloads and expire in hours or days. A CA, algorithm or key size can be changed by configuration. Cryptographic inventory covers algorithms and libraries, not only certificates, and post-quantum migration is under way. Lessons from industry incidents are turned into tested changes within a set time. |
Capability areas
Assess each of the six areas separately. The table shows what each looks like at Levels 1, 3 and 5; Levels 2 and 4 sit between them.
| Area | Level 1 Ad hoc | Level 3 Defined | Level 5 Optimised |
|---|---|---|---|
| Inventory and discovery | No list; certificates found when they expire | Continuous discovery across networks, cloud accounts, load balancers, Kubernetes and Certificate Transparency logs into one inventory, including private CAs | Inventory includes algorithms, key sizes, libraries and dependencies (a cryptographic bill of materials) and drives automated decisions |
| Ownership and policy | No owner field; anyone can request from any CA | Certificate policy approved; every certificate has an owning team and application; approved CA list enforced | Ownership derived automatically from workload identity and service catalogue; policy enforced as code at issuance |
| Issuance and automation | Manual CSR, email to CA, manual install | ACME or platform-native automation for all public TLS; enrolment protocols for devices; renewal tested | Issuance is a self-service API for every certificate type; no human handles private keys |
| Short-lived certificates and renewal | Longest available lifetime chosen to avoid work | Public certificates renewed at about two-thirds of lifetime; internal lifetimes set by policy and shrinking | Workload certificates live hours to days; renewal timing follows CA signals; outages from expiry do not occur |
| Key protection and revocation | Keys in files, copied between servers and into source control | CA and code signing keys in HSMs or managed key services; revocation procedure documented and tested; CT monitoring for unexpected issuance | Non-exportable keys for workloads; revocation replaced by short lifetimes where possible; compromise drills run |
| Crypto agility | No idea which systems depend on which CA or algorithm | Dependencies on each CA known; a documented plan to move off any CA; algorithm inventory started | CA or algorithm swap is a configuration change; replacement of all certificates from one CA proven within days; post-quantum pilots running |
Evidence by level
| Evidence an assessor can check | Level 2 | Level 3 | Level 4 | Level 5 |
|---|---|---|---|---|
| Inventory coverage versus discovery scans and CT logs | Partial, public only | All public and private CAs | Coverage measured and above target | Includes algorithms and libraries |
| Certificates with a named owner | Some | All in inventory | Owner verified on each renewal | Derived automatically |
| Public TLS renewed automatically | Some web servers | All | All, with success rate reported | All, with CA-signalled windows |
| Expiry-caused incidents in the last 12 months | Several | Rare, with root cause recorded | None in critical services | None |
| Signing and CA keys in hardware | Root CA only | All CA and code signing keys | Plus key ceremonies audited | Plus non-exportable workload keys |
| Mass replacement exercise | Never | Procedure documented | Exercised against a deadline | Proven within days for a whole CA |
Framework mapping
| Capability area | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 | MITRE ATT&CK threat addressed |
|---|---|---|---|
| Inventory and discovery | ID.AM-02, ID.AM-08 | CM-8 System Component Inventory | T1588.004 Digital Certificates (attacker-obtained certificates found via CT monitoring) |
| Ownership and policy | PR.AA-01 | SC-17 Public Key Infrastructure Certificates | Unowned certificates used for impersonation |
| Issuance and automation | PR.AA-01, PR.DS-02 | SC-17, IA-5(2) Public Key-based Authentication | Manual handling of private keys |
| Key protection and revocation | PR.DS-02 | SC-12 Cryptographic Key Establishment and Management | T1649 Steal or Forge Authentication Certificates; T1553.002 Code Signing |
| Crypto agility | ID.AM-08 | SC-12, SC-17 | Forced replacement after CA distrust or algorithm weakness |
In ISO/IEC 27001:2022 the programme sits mainly under Annex A 8.24 Use of cryptography, with inventory supported by 5.9 Inventory of information and other associated assets. NIST SP 1800-16 describes a formal TLS server certificate management programme for large and medium enterprises and is a useful reference for Levels 3 and 4.
How to run the assessment
- Set the scope: include public TLS, every private CA (including those built into cloud services, service meshes and Active Directory Certificate Services), code signing, client and device certificates, and SSH certificates if used.
- Gather evidence first: run discovery and compare it with Certificate Transparency results for your domains before interviewing anyone. The gap between what teams believe exists and what the scan finds is itself a finding.
- Score each area against the evidence table: award a level only when all its evidence is present; partial evidence stays at the level below.
- Take the lowest area as the overall level: report the area scores alongside so strengths are visible.
- Set a dated target: for most organisations, Level 3 across all areas before the 100-day limit applies on 15 March 2027, and Level 4 for public certificates before the 47-day limit on 15 March 2029.
- Repeat annually: and after any certificate-related incident or CA distrust event.
Self-assessment questions
Inventory and discovery
- Can we produce, today, a list of every certificate due to expire in the next 30 days, including those from internal CAs?
- Do we monitor Certificate Transparency logs for certificates issued for our domains that we did not request?
- Do we know every private CA running in our environment, including those inside cloud services and appliances?
Ownership and policy
- Does every certificate in the inventory have a named owning team and application?
- Is there an approved CA list, and is issuance from other CAs blocked or detected?
- Does the certificate policy set maximum lifetimes, key types and key storage per certificate type?
Issuance and automation
- What percentage of public TLS certificates renew without a human touching them?
- Can a developer get a correctly configured internal certificate through an API in minutes?
- Are private keys ever emailed, copied by hand or committed to repositories?
Short-lived certificates and renewal
- Will our renewal process still work when public certificates last 100 days, then 47 days?
- Do renewal jobs alert on failure early enough to fix them before expiry?
- Do our ACME clients support ACME Renewal Information so the CA can bring renewals forward?
Key protection and revocation
- Are all CA and code signing private keys held in HSMs or managed key services?
- When did we last test revoking and replacing a compromised certificate end to end?
- Would we detect a signing key that should have been retired but is still trusted?
Crypto agility
- If our main public CA were distrusted by browsers, how long would it take to replace every certificate it issued?
- Do we know which systems and libraries depend on RSA and elliptic curve algorithms that post-quantum migration will replace?
- Can we change CA or algorithm through configuration, or does it need code changes in each application?
Moving up: roadmap
- Level 1 to 2: run discovery across networks and cloud accounts, start CT monitoring, publish an approved CA list, and automate renewal for the highest-traffic public sites. See the Certificate Transparency Monitoring Guide for Web Teams.
- Level 2 to 3: adopt a certificate policy with owners, bring private CAs into the inventory, automate every public TLS certificate with ACME or platform-native tooling, move CA and signing keys into hardware, and document revocation. See the Certificate Management Policy Template and the ACME and Automated Certificate Management Guide.
- Level 3 to 4: report renewal success, coverage and time-to-replace as metrics; shorten internal lifetimes; adopt ACME Renewal Information (RFC 9773), which lets a CA suggest a renewal window and helps it handle mass revocation events; run a timed mass-replacement exercise. See the CA Compromise and Mass Certificate Replacement Playbook.
- Level 4 to 5: issue workload certificates on demand with lifetimes of hours or days, extend inventory to algorithms and libraries, and start post-quantum pilots. See the Cryptographic Inventory and CBOM Guide and the Post-Quantum Readiness Maturity Model.
Public CAs are already moving faster than the rules require. Let's Encrypt has announced a move from 90-day to 45-day certificates by February 2028, with a 6-day short-lived profile available by opt-in, and advises renewing at about two-thirds of a certificate's lifetime where a client does not support ACME Renewal Information.
Measuring progress
- Inventory coverage: certificates in inventory as a percentage of those found by discovery and CT monitoring.
- Ownership: percentage of certificates with a verified owner.
- Automation: percentage of public and of internal certificates renewed without human action.
- Expiry incidents: number of outages or warnings caused by expired certificates per quarter (target zero for critical services).
- Time to replace: measured time to replace all certificates from one CA, from the last exercise.
- Policy exceptions: number of certificates outside policy (lifetime, key type, CA) and their age.
Common failure patterns
- Automating the visible estate only: public web certificates are automated while internal CAs, appliances and client certificates continue to expire unnoticed.
- Inventory without owners: a tool lists thousands of certificates but nobody can say who to call when one expires or must be revoked.
- Longest lifetime by default: teams choose maximum validity to avoid renewal work, so the renewal process is exercised so rarely that it fails when needed.
- Keys that outlive their purpose: signing keys stay trusted long after they should be retired. In the Storm-0558 incident a 2016 consumer signing key that should have been retired in 2021 was used to forge tokens, and the US Cyber Safety Review Board found no system to alert staff to old, active signing keys. See the Microsoft Storm-0558 key breach analysis.
- No plan for losing a CA: mass replacement after a CA distrust event or a revocation deadline is improvised, with a five-day revocation window under the Baseline Requirements for many reasons and 24 hours for some. See CA Distrust Events.
Lessons from incidents
The Storm-0558 case shows why key protection and lifecycle belong in the same model as certificate renewal: the failure was not a missing certificate but a key that nobody retired, with manual rotation that had stopped after an outage. The lesson for each level is the same: automate rotation so that an outage does not become a reason to stop rotating, and make the inventory raise alerts on credentials past their intended life. Expiry outages teach the mirror lesson, covered in the Certificate Expiry Outages and 47-Day TLS Certificates Guide.
Practitioner checklist
- Run discovery across networks, cloud accounts, load balancers and clusters, and reconcile it with CT log results.
- Bring every private CA, including cloud and directory-integrated CAs, into one inventory.
- Assign an owning team and application to every certificate and verify it on renewal.
- Automate renewal for all public TLS certificates before the 100-day limit on 15 March 2027.
- Hold CA and code signing keys in HSMs or managed key services and audit key ceremonies.
- Monitor CT logs for unexpected certificates issued for your domains.
- Exercise replacing all certificates from one CA against a deadline, and record the time taken.
- Start an algorithm and library inventory as the foundation for post-quantum migration.
Standards and references
- CA/Browser Forum: TLS Baseline Requirements
- CA/Browser Forum: Ballot SC-081v3, Reducing Validity and Data Reuse Periods
- NIST: SP 1800-16 Securing Web Transactions, TLS Server Certificate Management
- IETF: RFC 9773 ACME Renewal Information (ARI) Extension
- NIST: CSWP 39 Considerations for Achieving Cryptographic Agility
- Let's Encrypt: Decreasing Certificate Lifetimes to 45 Days
- NIST: Cybersecurity Framework 2.0
Related NHI Mgmt Group resources: Machine Identity, PKI and Certificate Lifecycle Guide · Enterprise PKI Design and CA Hierarchy Guide · Top 10 PKI and Certificate Management Mistakes · NHI Governance Maturity Model