The acronyms are a mouthful. STIR — Secure Telephone Identity Revisited — is a set of IETF standards defining how to cryptographically sign a phone call's claimed originating number. SHAKEN — Signature-based Handling of Asserted information using toKENs — is the framework, developed by ATIS and the SIP Forum, that specifies how carriers actually deploy STIR in their networks, including the certificate governance that decides who gets to sign anything at all.
Together they answer a narrow question: did the carrier that put this call onto the network vouch for the number it claims to be from? Note what the question is not. It is not "is this caller honest," not "is this call wanted," and not even "is this the person who owns the number." Keeping that scope in mind explains both why the system works and why your phone still rings with scams.
How a Call Gets Signed
When a call originates on a SIP (internet telephony) network, the originating carrier creates a PASSporT — a Personal Assertion Token, defined in RFC 8225. Structurally it's a JWT, the same JSON-based signed token format used all over the web: a header naming the signing algorithm and the carrier's certificate, a payload containing the calling number, the called number, and a timestamp, and an ECDSA signature over the lot.
That token travels with the call in a SIP Identity header (RFC 8224). Every carrier the call transits passes it along. The terminating carrier — the one delivering the call to your phone — fetches the signing carrier's certificate, verifies the signature, checks the timestamp is fresh, and decides what to tell your phone. The certificates themselves aren't ordinary web PKI: they're issued under a closed governance system (RFC 8226 defines the certificate format) where a policy administrator decides which providers are eligible to sign calls at all. In the United States, the FCC designated a Secure Telephone Identity Policy Administrator to run that gatekeeping.
The signature doesn't authenticate the human caller. It authenticates a carrier's assertion about the calling number, and it makes that assertion non-repudiable. The real power of STIR/SHAKEN is traceability: when a scam campaign is signed, investigators know exactly which provider admitted it to the network.
The Three Attestation Levels
Not every carrier can fully vouch for every call, so SHAKEN defines three grades of confidence, embedded in the PASSporT:
| Level | Name | What the carrier is asserting |
|---|---|---|
| A | Full attestation | "This is my customer, and I know they're entitled to use this number." The strongest claim — typically your mobile carrier signing a call from your own phone. |
| B | Partial attestation | "This is my customer, but I can't verify their right to the number they're presenting." Common for enterprise PBXs that bring their own number blocks. |
| C | Gateway attestation | "I put this call on the IP network, but I have no idea where it actually came from." Typical for calls entering from international gateways or legacy equipment. |
Only A-level attestation supports the "Caller Verified" checkmark some phones display. B and C calls aren't blocked — they just carry less trust, and carriers' analytics engines weigh that when deciding whether to label a call "Spam Likely."
Why It Became Mandatory
The US Congress passed the TRACED Act in 2019, directing the FCC to require STIR/SHAKEN. Major US voice providers had to implement it on their IP networks by June 30, 2021, and Canada's CRTC imposed a similar mandate later that year. Alongside the signatures, the FCC built a Robocall Mitigation Database: every provider serving US numbers must register and describe how it prevents illegal traffic, and downstream carriers must refuse traffic from providers that don't.
That enforcement plumbing matters more than the cryptography. The FCC has since ordered carriers to cut off specific providers that persistently admitted illegal robocall traffic — something that was practically impossible when call origins were unprovable. Signed calls create an audit trail; the audit trail creates accountability; accountability is what actually changes provider behavior.
The Gaps: Why Scam Calls Still Reach You
If caller ID is now signed, why does your phone still ring with fake fraud-department calls? Several structural reasons:
- Legacy networks can't carry signatures. STIR/SHAKEN lives in SIP headers. Calls that traverse older TDM (circuit-switched) segments lose the token in transit, and smaller rural carriers were granted extended deadlines for exactly this reason. An unsigned call isn't necessarily fraudulent — so it can't simply be blocked.
- International calls arrive with weak attestation at best. A call from overseas enters through a US gateway that can only honestly assert level C. Most large-scale scam call operations originate abroad, which is precisely the traffic the system can say the least about.
- Verification ≠ legitimacy. A scam operation can buy real phone numbers from a legitimate VoIP provider and earn full A-level attestation. The signature then truthfully confirms a number the scammer genuinely controls. The checkmark says "not spoofed" — it does not say "not a criminal."
- The display layer is inconsistent. Whether you see "Caller Verified," a checkmark, or nothing varies by carrier and handset. Most users couldn't say what the indicator means, which limits how much fraud it can prevent.
A signed call is like a sealed envelope with a verifiable return address. The seal proves who mailed it. It tells you nothing about whether the letter inside is a birthday card or a ransom note.
What This Means for Your Threat Model
STIR/SHAKEN has genuinely reduced blatant number spoofing — the era when scammers freely impersonated your bank's published phone number or "neighbor-spoofed" your own area code is fading on IP networks. But the practical rules for an individual haven't changed:
Never authenticate an inbound call by its caller ID. Verified or not, an incoming call claiming to be your bank deserves a hang-up and a callback to the number on your card. Spoofing is harder now; social engineering is not, and voice cloning has made the voice on the line as forgeable as the number once was. Phone numbers also remain a weak anchor for identity in general — they can be SIM-swapped, and the signaling network behind them has its own decades-old vulnerabilities that signatures on caller ID do nothing to fix.
The deeper lesson is architectural. The phone network is retrofitting authentication onto an identity system it never designed, governed by thousands of carriers with uneven incentives — and the result, years into a federal mandate, is partial coverage and probabilistic trust. Systems designed with cryptographic identity from the start don't have this problem: in end-to-end encrypted messengers, the question "is this really the person I think it is" is answered by key verification between the endpoints, not by a chain of intermediaries vouching for each other. That's the difference between authentication as a property of the protocol and authentication as a patch.
Haven sits on that second philosophy: identities are cryptographic keys, and message authenticity is verified end-to-end rather than asserted by infrastructure. For voice and video, the same applies — encrypted calling inside an authenticated channel makes the "who is this really" question answerable in a way the public phone network, even with STIR/SHAKEN, cannot match. Use the phone network for what it's good at; just don't treat anything it displays as proof of identity.