Skip to main content

SMS Pumping

SMS pumping, also called SMS toll fraud or artificially inflated traffic (AIT), is a fraud scheme in which attackers make your application send text messages to phone numbers they earn money from. Bots enter those numbers into a form that sends a text, usually the verification code on a sign-up form. Your application pays to deliver each message.

The attacker isn't trying to take over an account or steal data. The message is the whole point, and your SMS bill is the payout.

How SMS pumping works

Every text message your application sends passes through a chain of companies. Your application hands the message to an SMS provider, the provider routes it through one or more aggregators, and the mobile carrier that owns the destination number delivers it. Each company takes a fee. The largest, the termination fee, goes to the destination carrier, and in some countries it's high enough that a single verification code costs $0.20 or more to send.

SMS pumping turns that fee into income:

  1. The attacker gets phone numbers that pay out. They partner with a carrier or aggregator that will share its termination fees, lease blocks of numbers under a revenue-share agreement, or sit inside the delivery chain themselves.
  2. They find a form that sends a text. Any public form works: sign-up, phone verification, password reset, or "text me a download link."
  3. Bots submit the numbers at scale. They request codes for thousands of numbers, often rotating through proxies so no single IP address stands out.
  4. Your application pays for every message. Each one is billed like a message to a real user. The destination carrier collects its fee and passes a share to the attacker.
  5. The codes are never used. Often there's no handset at the other end. The numbers exist only to receive billable traffic.

SMS pumping is a form of international revenue share fraud (IRSF), which has worked the same way with premium-rate voice calls for decades.

SMS pumping compared to similar attacks

  • SMS bombing floods one real person's phone with codes to harass them. The victim is the recipient, and the attacker earns nothing. In SMS pumping, the victim is the sender.
  • Credential stuffing and account takeover try to get into accounts. SMS pumping only needs the message sent.
  • Smishing is phishing by text message. The attacker sends the messages. In SMS pumping, you do.

SMS bombing and credential stuffing share many of the same defenses, because they also depend on automated requests to a public form.

Why sign-up forms are the target

A form that sends SMS passcodes is close to ideal for an attacker:

  • It's public. A sign-up form has to accept people you've never seen before, so it can't require users to sign in first.
  • It accepts any phone number. Unless you restrict it, that includes numbers in countries where you have no users.
  • It sends immediately and repeatedly. One request sends one message, and "resend code" sends another.
  • It works as designed. The attacker doesn't need a vulnerability. The form does what you built it to do.

This is why an application with a few hundred real users can receive a bill for tens of thousands of messages. The size of the attack depends on how fast bots can call the form, not on how popular the application is.

How to recognize an SMS pumping attack

Real users who ask for a code almost always enter it. During an attack, the verification rate (the share of sent codes that users enter) drops sharply. That drop is the clearest signal. Other signs usually appear with it:

  • A spike in SMS volume or phone sign-ups that doesn't match a launch, campaign, or press mention
  • Messages to countries where you have few or no users
  • Phone numbers that are sequential or share a long prefix, such as +XX 555 0100, +XX 555 0101, and +XX 555 0102
  • Repeated resend requests for the same numbers
  • Requests from data centers, proxies, or automated browsers
  • An SMS bill far higher than your user growth explains

What SMS pumping costs

Messages to high-fee destinations can cost 20 times what a domestic message does, so 100,000 pumped messages at $0.20 each is $20,000. Attacks often run overnight or over a weekend, when nobody is watching the volume.

The best-known case is Twitter. In December 2022, Elon Musk said the company was paying about $60 million a year for fake verification texts sent through 390 carriers. The figure is his own, and no one has independently confirmed it.

A flood of sends can also exhaust rate limits or sending quotas, so real users stop receiving their codes. Fake sign-up attempts inflate your funnel numbers, and if your application creates accounts before verification, unverified accounts pile up in your user list.

How to prevent SMS pumping

No single control stops SMS pumping, so combine several:

  • Send only to countries you serve. A country allowlist for SMS is the most effective single control. Pumping concentrates in a small set of high-fee destinations where most applications have no real users.
  • Stop bots before you send. Bot detection and CAPTCHA challenges only help if they run before the message goes out. After that, you've already paid.
  • Rate-limit by number, network, and device. Limit sends per phone number, per number prefix, per IP address, and per device or session. Add a cooldown between resends that grows with each request. See rate limiting.
  • Validate the number. Reject malformed numbers, landlines, and premium-rate ranges before sending.
  • Watch the verification rate. Alert when SMS volume to a country spikes or the rate falls.
  • Cap your spend. Set sending limits and billing alerts with your SMS provider, so an attack you miss stays small.
  • Offer methods that don't cost per message. Passkeys, authenticator apps, and email codes give attackers nothing to pump.

SMS pumping protection in Clerk

Clerk handles SMS verification for you, so several of these layers are built in:

  • Clerk blocks disabled countries. Numbers in countries you disable on your SMS allowlist can't receive SMS or sign up. Every application starts with a default set of countries enabled, so review the list instead of assuming it matches where your users are.
  • Clerk challenges suspicious sign-ups. Challenge suspicious sign-ups is on by default for new applications on every plan. It challenges sign-ups that look automated or come through VPNs, Tor, or proxies. It doesn't cover sign-ins.
  • Clerk rate-limits requests. Each IP address has a limit on how often it can start sign-ups and sign-ins, and Clerk's prebuilt components make users wait 30 seconds before requesting another code.
  • Clerk Protect can stop a send. Separately from the sign-up challenge, Clerk Protect can block or challenge an individual SMS send. Application Logs records these decisions as protect.sms.blocked and protect.sms.challenged events.
  • Clerk logs the messages it sends. The SMS logs view in the Clerk Dashboard shows the messages Clerk sent, their destinations, and their delivery status. It goes back as far as your plan's retention period, which is 1 day on the Hobby plan.

The Business plan adds more protections for applications under frequent attack, such as hourly sign-up limits per IP address, challenges for residential proxies, and blocks on VPNs, proxies, and hosting networks.

Clerk also supports passkeys, authenticator apps, and email codes, so you can turn off SMS entirely.

How to stop an SMS pumping attack

  1. Find where the messages are going. In the Clerk Dashboard, open the SMS logs view and look at the destination numbers. Pumped traffic clusters in a few countries and number prefixes. Messages you deliver through your own provider, with Delivered by Clerk turned off, aren't in this view. Check your provider's logs for those.
  2. Check both sign-up and sign-in. In Application Logs, look for sign_up.phone_number.verification.code_sent and sign_in.phone_number.verification.code_sent events to see which flow is sending the messages. If you enable phone-code sign-in, a sign-in request for an existing user's phone number also sends a text. Sign-up protections don't cover it.
  3. Disable the countries you don't serve. On the SMS page, select the Settings tab and disable every country where you don't have real users.
  4. Ensure sign-up challenges are on. On the Protections page, confirm that Challenge suspicious sign-ups is enabled. Older applications might still use legacy bot sign-up protection or have the protection turned off. On the Business plan, set Challenge to All until the attack ends.
  5. Disable the Native API if you don't ship a native app. Clerk's native SDKs don't support challenges yet, so attackers can use the Native API to skip them. Turn it off on the Native applications page.
  6. Contact support. If the attack continues, send Clerk support the affected application and the time the spike started.