Post-Quantum Cryptography Migration Checklist
If you’ve decided to start migrating to post-quantum cryptography and you want an ordered list of what to actually do — not another explainer on Shor’s algorithm — this is that list. It’s aligned to NIST’s migration model and the FIPS 203/204/205 standards (ML-KEM vs ML-DSA vs SLH-DSA), and it’s written for a security or IT lead running the project without a dedicated cryptographer.
Migration is not a rip-and-replace. It’s a phased program: inventory → prioritize → build crypto-agility → deploy hybrids → validate → monitor. Skipping the inventory is the single most common failure, because you can’t migrate cryptography you don’t know you have. Work the phases in order; within a phase, parallelize freely.
Phase 0 — Establish ownership and scope
- Name an accountable owner for the migration (usually the person who answers security questionnaires).
- Define scope: production systems, internal systems, third-party/SaaS, and long-retention data stores. Write down what’s explicitly out of scope.
- Identify your compliance forcing functions — the RFP, audit, HIPAA/SOC 2/CMMC review, or insurance renewal driving the timeline. Anchor the plan to that date, not to the arrival of a quantum computer.
- Decide your target standards up front: ML-KEM (FIPS 203) for key establishment, ML-DSA (FIPS 204) for signatures, SLH-DSA (FIPS 205) where a conservative hash-based signature is warranted.
Phase 1 — Build the cryptographic inventory
This is the load-bearing phase. The deliverable is a single document an auditor would accept.
- Public-facing TLS. Every external domain and endpoint: TLS version, cipher suite, negotiated key-exchange group, certificate signature and public-key algorithm. (Bootstrap it with the one-line handshake check.)
- Certificates. Enumerate all certs — public and internal. Record signature algorithm (RSA/ECDSA), key size, issuer, and expiry. Expiring certs are your cheapest migration opportunities.
- Key exchange in transit. VPN, SSH, service-to-service mTLS, database connections, message queues.
- Data at rest. Backups, archives, object storage, databases. For each, record the key-wrapping/key-establishment algorithm and — critically — the required confidentiality lifetime.
- Code signing / firmware / secure boot. Any signature that establishes software authenticity.
- Cryptographic libraries and versions. OpenSSL, BoringSSL, libraries in your application dependencies. Note which support post-quantum algorithms and which don’t — for OpenSSL, 3.5.0 is the first release with ML-KEM, ML-DSA and SLH-DSA.
- Hardware roots of trust. HSMs, TPMs, smartcards — these often have the longest replacement lead time. NIST IR 8547 specifically flags smart-card-based credentials as needing an early start because of multi-year issuance and replacement cycles.
Tag every entry as quantum-vulnerable (RSA, ECDSA, ECDH, DH, DSA) or quantum-safe / not-applicable (AES-256, SHA-256/384, HMAC). This tagging alone turns the inventory into a work queue.
Phase 2 — Prioritize by real risk
Not everything migrates at once. Rank each inventory item on three axes:
- Data-secrecy lifetime. How long must this stay confidential? Long-retention data (PHI, financial, legal, IP) scores highest — it’s already exposed to harvest now, decrypt later.
- Exposure. Public internet > partner network > internal-only.
- Blast radius. Root CAs, code-signing keys, and identity providers affect everything downstream.
Apply Mosca’s inequality: if (secrecy lifetime + migration time) > (time until a quantum computer exists), that item is already late — migrate it first. In practice, long-retention confidential data on public-facing systems is almost always the top tier.
Phase 3 — Build crypto-agility
You’re going to change algorithms more than once (hybrids now, pure PQC later, and possibly a different PQC algorithm if one is deprecated). Design for that.
- Route cryptographic operations through a central library or service, not scattered inline calls.
- Make algorithm choice configuration-driven, not hard-coded.
- Confirm your TLS-terminating infrastructure (load balancers, CDNs, reverse proxies) can be updated to negotiate post-quantum groups.
- Inventory anywhere key sizes are assumed — ML-KEM and ML-DSA keys and ciphertexts are substantially larger than RSA/ECC, and some protocols, buffers, or hardware have size limits.
Phase 4 — Deploy hybrid key exchange first
Hybrids give you post-quantum protection without betting everything on a young algorithm — the session is secure unless both the classical and the post-quantum part are broken.
- Enable hybrid TLS 1.3 key exchange on public endpoints:
X25519MLKEM768(IANA codepoint0x11EC), orSecP256r1MLKEM768(0x11EB) where the FIPS-validated component must be P-256 ECDHE (RFC 10024). See TLS 1.3 post-quantum for the mechanics. - Verify negotiation end-to-end after enabling — the hybrid key share makes the
ClientHellolarge enough that it’s typically split across two packets, which some middleboxes mishandle. - Check every TLS hop, not just the public one: CDN-to-origin, load-balancer-to-service, and service-to-database connections each negotiate their own key exchange.
- Roll out to internal TLS and VPN where supported.
- For data at rest, begin re-wrapping long-retention data keys with ML-KEM-based key establishment.
Phase 5 — Migrate signatures and PKI
Signatures are slower to migrate than key exchange because they touch PKI hierarchies and long-lived trust anchors.
- Pilot ML-DSA (FIPS 204) for new code-signing and internal PKI.
- Plan CA hierarchy changes — post-quantum or hybrid certificate chains — with your CA vendor.
- Use SLH-DSA (FIPS 205) where you need a stateless, hash-based signature with conservative security assumptions (e.g. firmware/root-of-trust signing) and can tolerate larger signatures.
Phase 6 — Validate and monitor
- Re-run the handshake check from an OpenSSL 3.5+ client and confirm the negotiated groups changed as intended. An older client can’t offer ML-KEM, so it reports classical key exchange even against a migrated server.
- Update the inventory to reflect migrated items — it’s a living document.
- Set up drift monitoring: reissued certificates, new services, and infra changes can silently reintroduce classical-only crypto.
- Keep the inventory + plan export-ready for your next questionnaire or audit. That artifact is your answer to the PQC line item.
Worked example: one inventory, three verdicts
Here’s how Phase 1 and Phase 2 play out on three hypothetical inventory rows for a company running a SaaS product behind a CDN.
| Inventory row | What the check found | Tier | Why |
|---|---|---|---|
app.example.com (customer data, retained for years) | Browser-to-edge: X25519MLKEM768. Edge-to-origin: X25519 | 1 — partly done | Long secrecy lifetime and public exposure. The edge is migrated but the origin hop is still classical, so the row isn’t done |
| Nightly database backups to object storage, kept 7 years under the company’s own retention policy | Data keys wrapped with RSA-2048 | 1 | Long-retention confidential data. Re-wrap under ML-KEM-based key establishment before anything at lower risk |
| Firmware signing key for a hardware appliance | ECDSA P-256, key in an HSM that doesn’t support ML-DSA | 2 — long lead time | Forgery needs a real-time quantum computer, so it’s not urgent in the harvest-now sense. But the HSM replacement takes procurement time, so start now |
The first row is the failure mode to watch for. The public handshake check passes, someone ticks the box, and the backend hop carrying the same customer data stays classical. Record each hop as its own inventory row and you won’t hit it.
The second row shows why secrecy lifetime outranks exposure. Backups never touch the public internet, but anything an attacker exfiltrates or intercepts in transit to storage stays readable for as long as RSA holds.
The fastest way to start Phase 1
The inventory is where most teams stall, so start there and start narrow. Your public TLS is the highest-signal, lowest-effort slice — it’s internet-visible, it’s what a customer’s security team will check first, and it maps cleanly to the NIST standards.
Start with the item you can finish today: check your public TLS handshakes and record the negotiated group and certificate algorithm for each domain. That’s the first checklist item done, and a defensible artifact to build the rest of the plan around.
Related: Post-quantum readiness self-check · NIST migration guidance · TLS 1.3 post-quantum key exchange
Frequently asked questions
What is a cryptographic inventory?
It is a list of every place your organization uses public-key cryptography, recording for each one the algorithm and key size, where it runs, what data it protects and for how long that data must stay secret. In a post-quantum migration it is the work queue: every row tagged quantum-vulnerable is a migration task.
What is the PQC migration timeline?
NIST IR 8547 (final, October 2026) deprecates quantum-vulnerable public-key algorithms such as RSA and ECDSA after 2030 and disallows them after 2035. Federal high value assets and high-impact systems must use post-quantum key establishment by December 31, 2030 and post-quantum signatures by December 31, 2031. For most private companies the effective deadline is whichever customer audit or contract adopts those dates first.
What are the biggest PQC migration challenges?
Hardware and trust infrastructure with long replacement cycles, larger keys and handshakes that break size assumptions in protocols and middleboxes, and cryptography nobody knows is there because it was never inventoried. NIST IR 8547 singles out PKI-based credentials such as smart cards and readers as needing an early start because of multi-year issuance cycles.
Do I need to replace AES in a post-quantum migration?
Generally no. Inventory it, but leave AES-128, AES-192 and AES-256 in place: NIST IR 8547 states it does not expect to transition away from its symmetric standards as part of the PQC migration. Only the handful of 112-bit symmetric standards need replacing, because NIST disallows them in 2030.
What is the NCCoE Migration to Post-Quantum Cryptography project?
It is a NIST National Cybersecurity Center of Excellence project, run with industry collaborators, that produces the SP 1800-38 practice guides. Those guides cover discovering where quantum-vulnerable cryptography is used and how to test and interoperate with post-quantum replacements, and they are useful as a model for your own inventory tooling.