Post-Quantum Firmware Signing: Roots of Trust in Long-Lived Government and Defense Equipment

Signature verification that is fixed in hardware at manufacture cannot be migrated later. What NIST IR 8547, CNSA 2.0, and EO 14412 say about code signing, how ML-DSA, SLH-DSA, LMS, and XMSS compare, and what to ask suppliers.

Federal post-quantum deadlines put key establishment ahead of digital signatures. The reasoning is sound for network traffic: a recorded key exchange can be attacked years later, while forging a signature requires a quantum computer at the moment of the forgery.

Firmware signing is the exception. A device checks its firmware with a public key and verification code that are often fixed when the device is manufactured. If that check uses RSA or elliptic-curve signatures and cannot be updated, the device will accept forged firmware for as long as it stays in service once those algorithms can be broken. For equipment with a twenty-year service life, the decision is made at design time, which for many programs is now.

What current guidance says about code signing

NIST’s final IR 8547, Transition to Post-Quantum Cryptography Standards (October 2026), treats code signing as its own use case. It notes that in some cases it is not feasible to update the code that performs signature verification after devices have been manufactured, and that such devices should be designed to require quantum-resistant signatures on executable code if there is a risk they will still be in use when a cryptanalytically relevant quantum computer exists. The same report deprecates RSA, ECDSA, and EdDSA signatures after 2030 and disallows them under NIST standards after 2035.

For national security systems, the NSA’s CNSA 2.0 suite put software and firmware signing first on its transition timeline: support and prefer the new algorithms by 2025, and use them exclusively by 2030. It named the stateful hash-based schemes LMS and XMSS for this purpose because they were already standardized.

For civilian agencies, Executive Order 14412 requires high value assets and high impact systems to use post-quantum digital signatures by December 31, 2031. Our post on federal PQC mandates covers the full set of dates. The order does not mention firmware specifically, so program offices have to recognize on their own that a 2031 signature deadline can imply a much earlier hardware decision.

Where a classical signature gets locked in

A boot chain usually has several verification steps, and they differ in how hard they are to change.

The first stage. Many processors start from a boot ROM that cannot be rewritten. It verifies the next stage against a public key, or the hash of one, programmed into one-time-programmable memory. Both the algorithm and the key are permanent. If the ROM only implements RSA or ECDSA, no software update changes that.

Later boot stages and the update client. Second-stage bootloaders, operating system loaders, and update agents are themselves firmware. They can be replaced with versions that verify post-quantum signatures, but only through an update that the device accepts under its current keys. That path stays open only while the old keys are still trustworthy.

The signing side. The hardware security module that holds the signing key, the build pipeline that calls it, and the certificate chain that vouches for the key all need to support the new algorithm. This side can be changed at any time, but it has procurement and validation lead times.

A cryptographic inventory built from network scans and certificate stores will usually miss the first two items, because boot verification never appears on a network.

Choosing a signature scheme for firmware

NIST has approved three families of post-quantum signatures, and they make different trades.

SchemeStandardSigner keeps stateApproximate signature sizeConsiderations for firmware
ML-DSA-87FIPS 204No4.6 KB (public key 2.6 KB)General purpose. Lattice-based, a newer security assumption. Named in CNSA 2.0.
SLH-DSAFIPS 205No8 KB to 50 KB depending on parameter setSecurity rests only on hash functions. Large signatures and slower signing.
LMS and XMSSSP 800-208YesA few KB, depending on parametersSecurity rests only on hash functions. Small public keys and compact verification code. Limited number of signatures per key. Named in CNSA 2.0 for software and firmware signing.

For comparison, an RSA-3072 signature is 384 bytes and an ECDSA P-384 signature is 96 bytes. Boot ROMs with tight size limits, and update channels with small bandwidth, need to be checked against the new sizes.

The stateful schemes need the most care on the signing side. Each LMS or XMSS private key is a set of one-time keys, and security depends on never using one of them twice. NIST’s project page for stateful hash-based signatures says they are not suitable for general use for that reason. SP 800-208 requires key generation and signing to happen inside a hardware cryptographic module that does not allow private keys to be exported, which complicates backup. NIST states that it is working on a revision that would permit export with mitigations. In the meantime, RFC 10033 (September 2026) describes how to handle state and backup, including the failure cases: restoring a signing module from a backup or cloning a virtual machine can silently reuse a one-time key and make forgery possible.

The program also has to budget signatures. A key that can produce a fixed number of them must cover every release, every variant, and every re-sign over the life of the fleet.

An example: a controller with a twenty-year service life

Consider a hypothetical program fielding an embedded controller, with units already deployed and a new production lot planned. All figures here are illustrative. The boot ROM verifies the second-stage bootloader with ECDSA P-384 against a key hash in fuses. The bootloader verifies the operating system image, and an update agent verifies packages, both with RSA-3072.

A review sorts the findings by what can still be changed.

  • New production lot. The first-stage check can be changed before manufacture, by selecting a processor variant whose ROM verifies LMS or ML-DSA, and by confirming the signing module supports the same scheme. This is the only point where the problem is fully solved.
  • Fielded units, later stages. A signed update replaces the bootloader and update agent with versions that require post-quantum signatures on everything they load. This narrows what a forged classical signature could reach.
  • Fielded units, first stage. The ROM check stays ECDSA. An attacker who could forge that signature would also need a way to write the first-stage image, such as physical access or a compromised update path. The program records this as a residual risk, adds compensating controls on physical and maintenance access, and schedules hardware replacement at the next depot overhaul.

The third finding becomes a written risk decision with an owner and an end date.

Questions for suppliers and program reviews

  • Which signature algorithm does each boot stage verify, and which of those stages can be updated in the field?
  • Is the root public key or its hash in one-time-programmable memory? Can a second key or algorithm be provisioned later?
  • Which hardware module holds the signing key, and what validation does it carry for the chosen post-quantum algorithm?
  • For LMS or XMSS: how many signatures does the key allow, how is state protected, and what is the recovery plan if the signing module fails?
  • Does the update protocol carry the larger signatures within its size and timing limits?
  • After a bootloader update, how do we confirm which units are enforcing the new check?

Evidence that the change reached the fleet

A signing pipeline can be switched to a new algorithm on one day. The fleet changes unit by unit, and a unit that missed the bootloader update still accepts the old signatures.

Sentinel helps organizations understand and change cryptography across many systems at once, including where signing keys and verification steps sit and what depends on them. CAM provides asset-level cryptographic evidence for assets in production after migration, so a statement such as “all fielded units require post-quantum signatures on updates” can be checked per asset. Both products have a provisional patent issued.

If your program is specifying new equipment or assessing a fielded fleet and wants to work through its signing and boot chain, contact QuantumWorks.

  • Firmware Signing
  • Code Signing
  • Secure Boot
  • Root of Trust
  • ML-DSA
  • LMS and XMSS
  • CNSA 2.0

Working on a hard technical problem?

We work with government organizations, research institutions, and technology teams across cybersecurity, software, and autonomous systems.