Q-Cyber Magazine

Defence · 8 November 2026 · 3 min read

Trust now, forge later: Defence authentication in the quantum transition

NSA's 'trust now, forge later' problem is the authenticity counterpart to harvest-now-decrypt-later: today's signatures and certificates must still be trustworthy when quantum-vulnerable algorithms fail.

Abstract graphic of a lattice of connected points in Knowlcom blue

Confidentiality gets most of the attention in post-quantum discussion. Defence also depends on a quieter property: the ability to know that an order, a software update or a device identity is genuine.

Authenticity has a half-life too

NSA's October 2026 account is straightforward. Modern systems use cryptographic algorithms to establish trust and to authenticate users and devices. Some of those algorithms, based on integer factorisation or discrete logarithms, are vulnerable to a future cryptanalytically relevant quantum computer. Once that capability exists, the mathematics that now protects authentication will no longer do so. The agency's name for the resulting pattern is "trust now, forge later."

This is distinct from harvest-now-decrypt-later. A recorded ciphertext can be opened later without any change to the original message. A forged signature, by contrast, is an integrity and identity failure: a software update that looks signed, a device that looks enrolled, a command that looks authorised. The two risks share a cause, but they demand different inventories.

NIST has already provided the replacement standards. FIPS 204 specifies a module-lattice digital signature scheme derived from CRYSTALS-Dilithium. FIPS 205 specifies a stateless hash-based scheme derived from SPHINCS+. NIST has also said it is developing a FIPS derived from Falcon as a further alternative, and that ML-DSA and SLH-DSA, with ML-KEM, should be put into use now as the foundation of most deployments.

Exhibit 1

Two quantum-era risks, two first standards

RiskWhat failsFirst-line NIST standardFederal high-value system marker (NIST IR 8547 path)
Harvest now, decrypt laterConfidentiality of recorded ciphertextFIPS 203 (ML-KEM) for key establishment31 December 2030
Trust now, forge laterAuthenticity of signatures, certificates and identitiesFIPS 204 (ML-DSA) and FIPS 205 (SLH-DSA)31 December 2031

Source: NSA, Post-Quantum Cryptography: A Digital Armor, 1 October 2026; NIST FIPS announcement, 13 August 2024; NIST IR 8547 announcement, 27 October 2026.

Reference

Size comparisonPublic keys: classical vs post-quantum
Public keys: classical vs post-quantum, in bytes
AlgorithmBarSize (bytes)
X25519 public key32 bytes
RSA-2048 public modulus256 bytes
ML-KEM-512 encapsulation key800 bytes
ML-KEM-768 encapsulation key1,184 bytes
ML-KEM-1024 encapsulation key1,568 bytes
ML-DSA-65 public key1,952 bytes

ML-KEM ciphertexts are 768, 1,088 and 1,568 bytes for the three parameter sets.

Sources NIST FIPS 203 (ML-KEM), FIPS 204 (ML-DSA), FIPS 205 (SLH-DSA), FIPS 186-5 (ECDSA); RFC 7748 (X25519); RFC 8032 (Ed25519).

Size comparisonSignatures: classical vs post-quantum
Signatures: classical vs post-quantum, in bytes
AlgorithmBarSize (bytes)
Ed2551964 bytes
ECDSA P-25664 bytes
RSA-2048256 bytes
ML-DSA-442,420 bytes
ML-DSA-653,309 bytes
ML-DSA-874,627 bytes
SLH-DSA-SHA2-128s7,856 bytes
SLH-DSA-SHA2-128f17,088 bytes

SLH-DSA keeps a 32-byte public key but carries the largest signatures.

Sources NIST FIPS 203 (ML-KEM), FIPS 204 (ML-DSA), FIPS 205 (SLH-DSA), FIPS 186-5 (ECDSA); RFC 7748 (X25519); RFC 8032 (Ed25519).

Where defence actually signs

The signature estate in a defence organisation is wider than e-mail. It includes certificate authorities, device enrolment, software and firmware signing, document authentication, and any protocol that binds a command to an identity. A migration that upgrades a few user certificates and leaves firmware signing on a vulnerable algorithm has not addressed the trust-now-forge-later problem.

The later date for signatures in the federal high-value path — the end of 2031, after key establishment in 2030 — should not be read as permission to delay the inventory. It reflects a sequencing of effort, not a judgement that authenticity is secondary. Certificate hierarchies and embedded signers often take longer to replace than session-key establishment, which is an argument for starting earlier, not later.

Implications

  1. Catalogue every signing purpose, including firmware, devices and commands, not only user certificates.
  2. Plan FIPS 204 and 205 adoption alongside FIPS 203, with hybrid or dual-signature designs where they reduce risk during transition.
  3. Protect the issuers. A post-quantum leaf certificate under a vulnerable certificate authority is a delayed failure.
  4. Rehearse revocation and re-enrolment. Forgery risk is also an incident-response problem.

What this means for defence

If a service cannot list the algorithms that underwrite its software updates and its device identities, it cannot yet claim a post-quantum plan. Encryption and authentication have to move together, on a published standard, before the old mathematics expires.

Sources

  1. NSA, Post-Quantum Cryptography: A Digital Armor
  2. NIST announcement of FIPS 203, 204 and 205
  3. NIST IR 8547 announcement