Summary

  • An active IETF draft proposes STIR certificates lasting days or hours and requires an embedded x5c chain when the lifetime is under one week. This can limit key exposure and avoid a certificate fetch during the call.
  • Certificate validity, telephone-number assignment, authority-token evidence, ACME authorization, SIP transport and final call treatment have different clocks. A short notAfter value does not make them one decision.

One valid chain, several changing facts

The opening is a constructed operating scenario, not a reported incident. Its purpose is to test the meaning of “fresh.” A verifier can establish that a certification authority issued a certificate for a defined interval, that its TN Authorization List covers the calling number, and that the private key signed the PASSporT. Those are valuable, bounded findings. They do not independently establish that the number assignment was checked at call time or remained unchanged after issuance.

draft-ietf-stir-certificates-shortlived-06, dated 6 July 2026, is now in IESG evaluation with the state AD Followup. The Datatracker records three DISCUSS positions and says two more YES or NO OBJECTION positions are needed to pass. It is intended for the Standards Track, but it is not an approved RFC and its text may change.

The proposal addresses a genuine operational trade. Traditional certificates may live for months or years. A short-lived certificate can expire in days or hours, reducing the useful window of a compromised signing key. If renewal is dependable, expiry can carry much of the freshness burden that otherwise falls on certificate revocation lists or online status checks.

Yet certificate time is only one part of STIR authority. A telephone number may be assigned, delegated or ported. A Token Authority may issue evidence that an ACME client presents to a certification authority. The authority evidence has its own expiry. The ACME account and challenge have their own state. The certificate then has issuance and validity times, the PASSporT has a creation time, SIP adds transit delay, and the terminating service applies its own verification and treatment policy.

Calling all of this “fresh” compresses a chain of evidence into one adjective. The safer record preserves which party asserted which fact, against which source, for what scope and until when.

The chain moves into the call

Revision 06 makes an important transport choice. If a certificate lasts less than one week, a compliant authentication service must place the certificate chain in the JWS x5c element, and the verification service must support it. The receiving side can therefore validate the path without fetching the certificate from the x5u location during the call.

The root should normally be omitted because the verifier already holds its trust anchors. The draft permits both x5u and x5c for compatibility, but they must identify the same chain. This prevents two headers from silently presenting different credential histories.

Carrying the chain by value changes failure location rather than abolishing failure. It removes a call-time certificate retrieval dependency. It also makes the PASSporT and SIP Identity header larger. A long chain or broad TN Authorization List can encounter intermediary header limits, UDP fragmentation or loss. A cryptographically valid assertion that does not reach the verifier intact cannot influence the call.

The privacy boundary shifts as well. Avoiding a live status or certificate query can reduce disclosure of calling relationships to an external service. A certificate scoped to one calling number can reveal less authority than a certificate covering a large range. But the chain now travels through the signaling path, where intermediaries may observe it. Whether the result is more private depends on scope, transport, caching and who can see signaling—not on lifetime alone.

Narrow scope does not synchronize authority

The draft allows a subject controlling many telephone numbers to request all, some or one of them. It even describes the possibility of a certificate issued for a single call, containing only the calling number and expiring shortly afterward. That is a powerful form of data minimization and consequence limiting.

It does not answer when the underlying number authority was last verified. RFC 9448 defines an Authority Token Challenge token for demonstrating telephone-number authority to the CA. The short-lived-certificate draft notes that this token might reasonably have a relatively long expiry so a client need not repeatedly return to the Token Authority. Operational efficiency and certificate narrowness can therefore coexist with older upstream evidence.

That is not proof of a defect in any deployment. A Token Authority may recheck assignments quickly, a CA may impose stricter policy, and operators may revoke or reject on new evidence. The point is that certificate duration cannot be used as a substitute for documenting those controls.

A defensible decision record joins the assignment source and observation time; the authority token, issuer and expiry; the ACME account and challenge; the CSR’s requested scope and duration; the certificate serial, issuance, validity and TN Authorization List; the signing key; the PASSporT creation time; the chain actually carried; the verifier policy; and the final call treatment. Without those links, investigators know only that one cryptographic interval was satisfied.

Renewal becomes part of call availability

Short lifetimes transfer work from the terminating verifier to the signing operator. The verifier can avoid a synchronous lookup, but the signer must obtain new certificates continually. CA reachability, ACME authorization, account integrity, clock accuracy and key custody become production dependencies for authenticated calling.

Renewal at the last possible moment is an outage design. Operators need a lead-time budget that covers CA latency and transient error, staggered renewal across fleets, bounded retries with backoff, capacity to overlap old and new credentials, and an explicit policy for what happens when the next certificate is unavailable. ACME STAR offers a way to pre-issue renewed credentials and reduce the client’s synchronous acquisition burden, but its schedule and delivery channel remain operational dependencies.

Monitoring should distinguish acquisition from use. Track time remaining at successful renewal, failed challenges by reason, CA and Token Authority reachability, unexpected changes in requested scope, certificate-chain size, SIP rejection caused by header limits, signature and path failures, expiry failures and downstream call treatment. A certificate dashboard that reports only “valid” hides the very transitions the mechanism makes frequent.

The shortest lifetime is not automatically the safest. A shorter interval reduces compromised-key exposure but also shrinks recovery margin and increases issuance volume. Local operators should choose the interval from their threat model, authority-update cadence, renewal capacity, transport limits and outage tolerance.

Verification informs treatment; it does not command it

STIR verification answers whether an identity assertion satisfies cryptographic and credential policy. The terminating service still decides how to treat the call. It may complete, label, challenge, divert or block based on additional evidence and local obligations. A valid short-lived certificate must not become an automatic command to complete a high-risk call, just as a transport failure should not be mistaken for proof of impersonation.

This separation matters during exceptions. If the signing operator cannot renew, the receiving network may see an expired credential, no credential, or an assertion that was stripped by an intermediary. The record should distinguish each case. Local policy can then balance fraud, availability, emergency calling and customer harm without falsifying the cryptographic result.

The draft is strongest as a narrow coordination mechanism. It defines how a bounded credential may be issued and carried. Trust-anchor choice, authority evidence, renewal margin, signaling transport, privacy trade-offs and call treatment remain local decisions. Running evidence—not a “short-lived” label—shows whether those decisions work together.

Sources