Why avoid storing biometric templates?

There is a house style for quantum security copy and it works by making the reader nervous. Nothing useful is bought with that. What an evaluator needs is the narrow version: which cryptography a large quantum computer actually defeats, what the standards bodies have already published in response, and which holdings are exposed in a way that survives scrutiny rather than in a way that merely sounds serious.

Those facts are settled and unglamorous. The machine does not exist yet. The replacement algorithms do: NIST published FIPS 203, 204 and 205 in August 2024, and US NSM-10 put 2035 on the calendar as the migration goal. What remains is sequencing and procurement rather than research.

Urgency, where there is any, comes from the recording rather than from the machine. Anything captured now under encryption that will not outlast the secret inside it is a live problem today, and the number of categories that genuinely qualify is small. A biometric qualifies without argument.

The argument then moves off strength and onto absence. Controls are bought to protect a store; it is the store that gives a recording its value; take the store away and a patient adversary is recording against nothing, however capable the hardware eventually becomes.

The triage is then run on our own architecture without softening. A deployment here is biometric-blind, so there is no face database behind the API and no template vault behind the console. The paper assumes a full-take recording over years, assumes the classical key exchanges eventually fall, and works through what the adversary is holding, artifact by artifact, once the plaintext is finally in hand.

The quantum-safe claim here describes the face-token construction. It uses NIST 140-3 approved symmetric and hash primitives exclusively: AES-256-GCM, HKDF-SHA256 and SHA-256. No proprietary or non-approved cryptographic primitives are used. The token contains no public-key structure for Shor’s algorithm to attack. PKI and IdP federation have separate algorithm and key lifecycles.

1. Which primitives Shor's algorithm ends and which Grover's algorithm merely weakens, and why that split decides what to migrate first
2. How Mosca's inequality behaves when the secret in question stays valid for as long as its owner does
3. The full-take recording scenario worked through artifact by artifact, first at capture and then after a hypothetical future decryption
4. The primitives inside the face token, and the exact sense in which the construction earns the description quantum-safe face token
5. Why short-lived signature artifacts are the wrong target for a decrypt-later strategy, and where hybrid post-quantum TLS (X25519MLKEM768) already applies today
6. Five tests an operator can apply to whichever biometric system they already run, none of which depend on this platform

Risk leads and cryptographers assembling a post-quantum inventory, and the reviewers who will have to defend that inventory to somebody else. Every claim here is narrow enough to be tested, which is the only kind worth entering in a register.

It is also useful pointed outward. Vendors have begun describing entire platforms as quantum-safe, and the paper gives a reader the language to ask which construction the label actually covers and what the protocol layer around it is doing in the meantime.

Seven pages, including an artifact table for the capture-and-decrypt scenario and a timeline that puts the published standards, the migration window and the credible machine estimates on one axis. The checklist at the end is the part to carry into a planning meeting.

The gated download sits on sensecrypt.com, and a copy of the PDF is available on request. If your own inventory work raises something the paper does not cover, that is the more interesting conversation to have with our engineers.