M9 / Key management and HSMs

Key ceremonies and KMIP

PractitionerAdvisor

After this lesson you can

  • List what concretely changes in a key ceremony when the root key is ML-DSA (backups, fingerprint verification, entropy)
  • Present the current status of KMIP's PQC support (a draft, not yet an approved standard) in the right frame

Before thisHSM validation reality: CAVP vs CMVP

Mental model

A key ceremony is a procedure in which a root CA key is generated in front of witnesses, following a strict script. In the move from RSA to ML-DSA the general structure of this procedure (witnesses, script, video recording, M-of-N control) does not change; what changes are the physical and time assumptions of some steps inside the script.

What concretely changes

Key and signature size affect physical media planning. An ML-DSA-65 public key is 1952 BSOURCED, many times a comparable-strength ECC key (64 BDERIVED). A ceremony script usually includes steps for transcribing key fingerprints by hand or writing them to backup media (smart cards, USB tokens); the capacity and time assumptions of these steps must be reviewed before being copied from old scripts written for RSA/ECC.

Entropy consumption. Lattice-based key generation (ML-DSA, ML-KEM) consumes more randomness per operation than ECC key generation. Making sure an HSM’s random bit generator (RBG) can meet that demand is a new item to add to the pre-ceremony checklist.

The general structure of the ceremony does not change. Witness requirements, M-of-N access control, video recording and script approval stay the same; this is not a “PQC changes everything” story but a story of updating a specific, limited set of physical and operational assumptions.

Backup and export procedures. The step of exporting a key from an HSM by splitting it into shares (M-of-N secret sharing) is also affected by the size difference: the physical format each share will be written to (paper, smart card, USB token) may have been designed for RSA-4096’s key of a few hundred bytes; the same format and capacity assumptions must be tested in advance for ML-DSA-87’s much larger private key. This is another concrete example of why a ceremony script cannot be updated by “copy and paste”; whether each step really works with PQC sizes must be verified with a rehearsal before the real ceremony.

KMIP: progressing, but still a draft

KMIP (Key Management Interoperability Protocol) is a standard language for HSMs and key management systems to talk to each other. For PQC support, KMIP v3.0 defines two new operations (Encapsulate and Decapsulate, the same logic as the new functions in PKCS#11 v3.2) and adds a KEM Algorithm enumeration (including the ML-KEM-512/768/1024 variants). But unlike PKCS#11 v3.2 in the previous lesson, KMIP v3.0 has DRAFT2026-07-14 status: it is still at the Committee Specification Draft (CSD02) stage, not a fully approved OASIS Standard.

The difference matters: telling an architect “KMIP already supports PQC” carries a different certainty claim from “the draft of KMIP’s next version defines PQC support, but it is not final yet”. When planning key management system integration, the uncertainty about when this draft will be final must be carried openly; it is the “draft and final are different” discipline you see throughout this course, applied to KMIP.

Numbers to know

  • An ML-DSA-65 public key is 1952 B, many times a comparable-strength ECC key (64 B); this difference means ceremony scripts and backup media capacity planning must be reviewed in advance
  • KMIP v3.0 (PQC support: Encapsulate/Decapsulate operations, KEM Algorithm enumeration) is still a draft (CSD02); unlike PKCS#11 v3.2, it is not a fully approved OASIS Standard

Lab: A PQC review checklist for a ceremony script

[not run] This is a checklist exercise, not a runnable command

Requires: . Check your setup

shell
# Take an existing RSA/ECC ceremony script and, for each step, ask: does this step's capacity or time assumption still hold with ML-DSA's large key and signature size?
Recorded output
Usually three steps are problematic: the fingerprint transcription format, backup media capacity, and the RBG/entropy budget (see the three sections above)

At the table

How to say this in a bank meeting.

To an executive
When moving our root key to ML-DSA, we are reviewing not the ceremony itself but the physical assumptions of the ceremony script (paper format, media capacity); this will be verified with a rehearsal before the ceremony.
To an architect
In planning our KMIP integration, we keep in mind that KMIP v3.0's PQC support is still a draft (CSD02) and that we should not tie our production systems to a non-final specification.
Objection
“"KMIP already supports PQC. Let's plan our integration around it now."”
Answer
KMIP v3.0's PQC support is real and progressing, but it is still at the draft (CSD02) stage, not a fully approved standard like PKCS#11 v3.2. Rather than locking the integration architecture today, we should keep tracking the final form of the draft.

Sources

  • Encryption Consulting, 2026. The only practitioner source found that describes the concrete practical effects of PQC in a key ceremony (media capacity, the fingerprint transcription process, entropy consumption)

    "What Changes in the Key Ceremony for PQC Key Generation" section / 10 min

    Commercial stake: Encryption Consulting is a PQC consultancy; the content looks accurate and specific, but it should be noted that it is a commercial source

  • OASIS KMIP Technical Committee, 2026. The source of the current, official status of KMIP's PQC support (Encapsulate/Decapsulate operations, KEM Algorithm enumeration)

    announcement text, status information / 5 min

Checkpoint

Answer first, then compare with the model answer and score yourself against the rubric. Saved in this browser only.

  1. 01Recall

    When the root key is ML-DSA, what concretely needs to be reviewed in a key ceremony?

  2. 02Recall

    What status does KMIP v3.0's PQC support have: a draft or a fully approved standard?

  3. 03Scenario

    A bank plans to reuse its existing RSA-4096 root ceremony script unchanged for ML-DSA-87. Say which assumptions must be re-checked.

  4. 04Hostile

    An architect says 'KMIP already supports PQC; let's plan our integration around it.' Which part of this claim is right and which is premature?

Project linkThe source of the PQC-specific ceremony script review steps in Project 2's (PQC PKI) rollover runbook.