Skip to content
By 11 min readMFA

Your Phone Number Is Not an Authenticator: Why Carrier-Only OTP Is Failing Enterprise Security

Blocking VoIP numbers stops cheap fraud at signup. It does nothing for account takeover, and it moves your security perimeter into a telco call center you don't control. The threat model, the regulatory direction, and a 90-day migration plan.

Your Phone Number Is Not an Authenticator: Why Carrier-Only OTP Is Failing Enterprise Security, by Deepak Gupta on guptadeepak.com

Open any major bank, insurer or brokerage app in North America and try to register a Google Voice number. You will get some version of "Please use a mobile number." Many of these companies then make that same mobile number the only way to log in, recover the account, or pass a call-center check.

Blocking non-fixed VoIP is a reasonable fraud signal. Making a carrier-delivered text message the only authenticator is a design failure, and regulators have started to say so. This piece covers the threat model, where the rules are heading, the lockout cost most teams never measure, and a 90-day plan to stop SMS being the last line of defense.

The VoIP ban is a fraud control dressed up as authentication

I spent years building a customer identity platform that grew past a billion managed users. The pattern here is familiar: a control designed for one job slowly gets stretched to do three. The phone number is being asked to prove who you are at signup, prove you are present at login, and prove you own the account when everything else fails.

It is good at the first job, mediocre at the second, and dangerous at the third.

One number, three jobsThe same phone number is asked three different questions
SignupIs this a real, established person?Good signal
LoginIs that person here right now?Phishable, interceptable
RecoveryDoes this person own the account?Whoever holds the SIM wins
Bars show how well a code sent to the number answers each question, as an editorial judgement, not a measured score.

Why enterprises block VoIP: the number as an identity anchor

The logic is sound as far as it goes. A postpaid mobile line comes with carrier records, a billing address, and years of history. A non-fixed VoIP line can be provisioned by API in seconds with no identity behind it.

Fraud teams see the difference in their data, and the tooling makes the filter easy. Twilio Lookup returns a line type, including nonFixedVoip, before you ever send a code, and many verification flows simply refuse that type.

So the VoIP filter answers a real question: is this a real, established person signing up? That is identity proofing. Even there, NIST treats a phone account as weak evidence: in SP 800-63A-4 it counts only as "fair" evidence.

The mistake happens one step later. The number that helped show the person exists is reused as proof that the person is present, every time they log in, for years. Those are different questions, and the phone network answers the second one badly.

The threat model: six ways a carrier-only OTP fails

When SMS is the only factor, your attack surface includes every system and every person between your OTP gateway and the customer's screen.

Where the code can be takenEvery hop between your server and the customer is an attack point
Your OTP serviceGenerates the codeReal-time phishing relay
SMS aggregatorRoutes to the carrierThird party in the path
Carrier networkSS7, Diameter, coreInterception
Carrier store and supportDecides which SIM owns the numberSIM swap, port-out
Your call centerThe recovery pathSocial engineering, cloned voice
Blocking VoIP at signup touches none of these five points.
Attack How it works Does blocking VoIP help?
SIM swap Attacker convinces the carrier to move the number to a new SIM or eSIM No
Port-out fraud Attacker opens an account at another carrier and ports the victim's number No
Network interception Attacker with access to SS7, Diameter or a carrier core reads SMS in transit No
Real-time phishing A fake login page relays the OTP to the real site within seconds No
Call-center social engineering An agent resets the phone number or bypasses OTP after a convincing story No
Synthetic voice A cloned voice passes voice biometrics or fools an agent No

Three of these deserve emphasis.

The carrier core is not a trusted zone anymore. After the Salt Typhoon intrusions into U.S. carriers, CISA's mobile communications best practice told highly targeted individuals to stop using SMS as a second factor, because SMS messages are not encrypted. The guidance was scoped to high-value targets. The weakness is not.

SIM swap is a social engineering problem, not a cryptography problem. In November 2023 the FCC adopted rules requiring U.S. carriers to authenticate customers securely before a SIM change or port-out and to notify them immediately. The compliance date for those rules is still pending, and even when they apply, the notice arrives after the request. Your security now depends on another company's store staff.

Voice is no longer proof of anything. Pindrop's 2025 Voice Intelligence and Security Report found deepfake activity up 680% in 2024 across the calls it analyzed. If your call center accepts "read me the code we just texted you" plus a familiar-sounding voice, you have two weak factors, not one strong one.

The same pattern shows up in workforce attacks: when attackers can't phish a passkey, they target the enrollment and recovery step instead. Customer recovery is that step at much larger scale. The Scam Atlas SIM swap entry collects real cases.

Where regulators are heading

North America is the outlier. Elsewhere, regulators have moved past "mobile versus VoIP" and are asking banks to stop depending on SMS altogether.

Date Regulator What changed
Apr 2026 Reserve Bank of India Authentication directions for digital payments apply: two distinct factors, one of them dynamic for most payments. No factor is mandated, so banks may move beyond SMS OTP
Mar 2026 Central Bank of the UAE Deadline to finish moving customers off SMS and email OTPs; the phase-out began in July 2025
Jul 2025 NIST SP 800-63B-4 final: codes sent by SMS or voice over the phone network are its one "restricted" authenticator
Dec 2024 CISA Tells highly targeted individuals to stop using SMS as a second factor
Jul 2024 Monetary Authority of Singapore Major retail banks phase out SMS OTP for logins by customers with a digital token
Nov 2023 FCC SIM swap and port-out rules adopted: secure carrier authentication plus immediate notice. Compliance date still pending

The NIST change matters most for U.S. enterprises. "Restricted" does not mean banned. Under SP 800-63B-4, if you keep it you must offer at least one unrestricted alternative, give users meaningful notice of the risk and the alternatives, assess the risk, and record a migration plan. If SMS is your only factor, you can't meet the first condition by definition.

My prediction, as analysis rather than reporting: U.S. and Canadian bank regulators will not issue a UAE-style ban. Liability will move instead. The first large court ruling that holds a bank responsible for SIM-swap losses because it offered no alternative to SMS will change roadmaps faster than any guidance document.

The cost nobody models: lockout

Security reviews count fraud losses. They rarely count the customers who simply cannot get in.

Carrier-only OTP fails legitimate users in predictable ways:

  • Travelers. Roaming is off, the home SIM is swapped for a local eSIM, or a short-code text may not arrive on a foreign network or over Wi-Fi calling.
  • Number changers. People change carriers, move countries, or lose a prepaid number after inactivity. Number recycling can hand their old codes to a stranger.
  • Small businesses and remote workers. Many run on VoIP lines by design and now can't use them for the accounts that matter most.
  • Call-in customers. The phone channel often asks for the same SMS code, so a customer who can't receive texts is locked out of both channels at once.
The lockout spiralA real customer who can't receive a text becomes your riskiest ticket
1Code doesn't arriveRoaming off, new eSIM, recycled number, VoIP line.
2Customer calls supportThe phone channel asks for the same SMS code.
3Agent makes an exceptionKnowledge questions, a story, a familiar voice. The agent is trained to help.
4Attackers learn the scriptThe exception path is the one an attacker with a convincing story uses too.
Every lockout trains your contact center to say yes to people who can't prove anything.

Recovery is where attackers aim, because it is built to say yes to people who can't prove anything. In my experience, the honest measure of an authentication system is not its login success rate. It is how it treats a real customer who has lost their factor, and whether an attacker can tell that same story just as well.

The reference architecture: signal, not proof

The fix is to separate two jobs the phone number has been forced to do. Keep it as a risk signal everywhere. Stop treating a code sent to it as proof anywhere it matters.

Reference architectureThree stages, one signal layer underneath
1. EnrollPrimaryPasskey on the device in handAlso offerPush approval in your app, recovery codes, a second deviceSMS: verify the number, not the person
2. Sign inPrimaryPasskey or trusted deviceStep upIn-app approval with number matchingSMS: fallback only, low-risk actions
3. RecoverPrimarySecond enrolled factor or recovery codeContact centerCalls started inside the app, passphrase, voice-clone checksSMS alone: never
Phone intelligence, checked before every send and every sensitive actionLine typeSIM changed recentlyNumber portedCall forwarding onNumber tenureDevice seen before
Low risk: continue with the method the customer enrolledHigh risk: step up to a passkey or app approval, or hold for review. Don't send the code.

Phone intelligence runs underneath all three stages and decides when to step up. The authenticator itself is a passkey or a device-bound app approval, and recovery becomes the hardest door, not the easiest. Silent network checks, which confirm the number with the operator without a code, are covered in my guide to silent network authentication.

The vendor landscape, by layer

No single vendor fixes this. The right stack pairs phone intelligence, so the number stays useful as a signal, with phishing-resistant authenticators, so it stops being the proof, plus a hardened contact center.

Layer Examples What it does for this problem
Phone risk intelligence Prove, Twilio Lookup, Telesign Line type, SIM tenure, porting and SIM-change signals, so you can skip the code and step up when risk is high. Prove says nine of the top ten U.S. financial institutions use it; Twilio's SIM swap and call-forwarding checks are in limited beta
Carrier network APIs GSMA Open Gateway through aggregators such as Vonage and Infobip Standard CAMARA APIs for SIM Swap and Number Verification, straight from the operator. GSMA counts 86 operator groups covering 80% of mobile connections
Passkeys inside a CIAM platform Okta (Auth0), Ping Identity, Microsoft Entra External ID Passkeys, push MFA, adaptive step-up and recovery flows in one customer identity platform
Passkey layers over an existing login Authsignal, Corbado, Twilio Verify Drop-in passkey and step-up when you can't replace the core identity provider
Hardware keys Yubico, Google Titan Phishing-resistant FIDO2 keys for high-value customers, executives and agents
Contact center Pindrop Device and network signals plus synthetic-voice detection, so agents stop depending on "read me the code"

Two cautions from building in this market.

First, phone intelligence makes SMS safer, not safe. A risk score can tell you a SIM changed last night. It can't stop a real-time phishing page from relaying a valid code.

Second, passkey adoption is now real enough to plan around. The FIDO Alliance estimates 5 billion passkeys in use. I track enrollment and sign-in share in my FIDO Passkey Index 2026 notes.

A 90-day migration plan

You don't need to rip out SMS on day one. You need to stop it being the only door. For the replacement order in more depth, see the CIAM Compass guide to deprecating SMS OTP.

90 daysFrom "SMS is the only door" to "SMS is a fallback"
30See the riskSIM-change and line-type checks before every send. Measure lockouts by country.
60Add a better doorPasskeys, in-app approval, trusted devices, recovery codes, a second factor.
90Close the side doorsNo number change on SMS alone. Calls start in the app. Risk assessment updated.
  1. Days 1 to 30: see the risk.
    • Add line-type and SIM-change checks before every OTP send, at login and at recovery, not only at signup.
    • Route high-risk sends (recent SIM change, recent port, forwarding on) to step-up or manual review instead of texting the code anyway.
    • Measure lockout: tickets, abandoned logins and recovery calls caused by SMS failures, split by country.
  2. Days 31 to 60: add a better door.
    • Offer passkeys in the web and mobile app, and turn on in-app push approval with number matching for existing app users.
    • Mark enrolled devices as trusted, so travelers can sign in from them without a fresh code.
    • Give customers single-use recovery codes and let them register more than one factor.
  3. Days 61 to 90: close the side doors.
    • Require a strong factor or an in-app approval to change a phone number, never SMS alone.
    • In the contact center, start authenticated calls from inside the app, add a customer-set verbal passphrase, and add synthetic-voice detection.
    • Publish a customer page explaining the options, and update your risk assessment to match NIST's "restricted" status for SMS.

Internally, track one number above the rest: the share of high-value actions (payee changes, limit changes, number changes) completed without SMS. That share is your real exposure.

The question worth asking your team

Blocking VoIP was never the problem. Stopping there was.

If an attacker ported your CEO's phone number tonight, how many of your customer-facing systems would let them reset an account before breakfast? If the honest answer is "most of them," your authenticator is the telco's call center.

What is the one flow in your product where SMS is still the last line of defense?

FAQ

Why do banks reject Google Voice and other VoIP numbers? Non-fixed VoIP numbers can be created instantly with no identity checks and recycled after use, so fraud teams treat them as high risk at account opening. The filter works against fake accounts, not against account takeover.

Is SMS OTP still allowed under NIST SP 800-63B-4? Yes, as a "restricted" authenticator. Organizations that use it must offer at least one unrestricted alternative, tell users about the risk, assess it, and plan a migration.

Is a mobile number safer than a VoIP number for MFA? For signup fraud, usually yes. For login security, not by much. Both receive unencrypted, phishable codes, and mobile numbers add SIM swap and port-out risk.

What should replace SMS OTP? Passkeys or in-app push approval bound to a registered device, with phone-risk signals used for step-up decisions, and recovery that never relies on SMS alone.

How do we protect the call center? Start authenticated calls from inside the app, use device and network signals, add synthetic-voice detection, and never let an agent change a phone number based on an SMS code alone.

Related reading: implementing FIDO2 authentication, step by step and the phishing-resistant authentication glossary entry.

Get new Identity & CIAM writing

Enjoyed this? Subscribe and tell us what you read most. Identity & CIAM is already ticked for you. No tracking pixels, unsubscribe with one click.

Tell us what you read most (optional)

About DeepakPublicationsAnalysisAll tracks