All articles Fundamentals

Artificially Inflated Traffic SMS: A Control Guide

Developer answering an OTP request at a dark desk lit by blue monitor light.

Artificially inflated traffic (AIT) in SMS OTP is a billing attack, not an account takeover. Someone scripts your ‘send code’ endpoint and points it at a list of phone numbers, and every request you answer becomes a message you pay for. The accounts are a side effect; the invoice is the target.

Most fraud controls are built for intent — is this person who they claim to be? AIT produces requests that look almost identical to real ones: a number and a tap. You can’t separate them by behavior alone, so you read the context around the request and answer it one request at a time.

This guide covers the economics, the four signals that show up before the bill does, and how to answer each OTP request with allow, step-up or block. Prynt Control makes the decision; your server still sends the code. Prynt does not send SMS.

Why AIT is an economics problem first

An AIT campaign is arithmetic. The attacker’s cost per request is close to zero: a script, a proxy list, a block of numbers. Their return depends on two things you control — what you pay per message and where the message terminates. When part of that termination fee flows back to the sender, volume becomes revenue.

That is why the usual defenses miss:

  • Per-IP rate limits fail when requests arrive through rotating proxies and datacenter ranges.
  • Per-number caps fail when the attacker buys a fresh block of numbers.
  • Per-account caps fail when the request carries no account at all, only a number.

You are not defending an account. You are defending a budget line, which means the policy has to be about devices, destinations and volume.

The signals that show up before the invoice

Analyst watching OTP request cost lines glow on a dark glass wall.

One device asking for codes for many numbers

Underneath a VPN hop or a cleared cookie, the device often stays the same. One fingerprint asking for codes for dozens of numbers in an hour is not a shared family phone. Count the distinct numbers per device, and cross-check accountsOnDevice and linkedId so a household with three logins is not treated as a farm.

Sequential numbers inside one block

Purchased blocks tend to be consecutive: same prefix, last digits stepping by one. Numbering plans do hand sequential ranges to real carriers, so treat this as a strong signal rather than proof — much stronger when the sequence shares a device, a proxy range or a short time window.

A spike that doesn’t match the funnel

Raw OTP volume tells you little on a good marketing day. Track ratios instead: OTP requests per signup, per login, per minute. AIT appears as a ratio that breaks away from its normal band while sessions and conversions stay flat. Volume at 03:00 local time for a destination you do not sell to is the same signal from another angle.

High-risk destinations and mismatched geography

Compare the number’s prefix and country against the request’s IP geolocation and the account’s billing country. VPN, proxy, Tor and datacenter signals belong in the same check. A US IP, an Indonesian number and no account history is a different request from your returning user’s.

From signals to one decision per request

Signals are the easy half. The useful half is a policy that returns something your code can act on: allow, step-up or block, with a reason code and a suggested step.

DecisionWhenWhat your server does
allowDevice and destination match the account’s historySend the code
step-upMixed signals: new device, proxy, first-seen numberRequire a captcha or email verification before sending
blockSequential block, many numbers on one device, destination riskReturn a soft error and log the reason code

Pseudo-code for the shape of it:

const d = await control.decide({ useCase: "sms", phone, visitorId, requestId });
// d.decision: "allow" | "step-up" | "block"
if (d.decision === "allow") return sendCode(phone);
if (d.decision === "step-up") return askForCaptcha(d.suggestedStep);
return softFail(d.reasonCode);

Two rules keep this honest. Prynt never blocks on its own; your code owns the verdict. And step-up should be the default for uncertainty — a captcha costs a real user a few seconds, a wrong block costs a customer.

Watch it before you enforce it

Start in mirror mode. Control’s 14-day mirror period records what it would have decided for each request without blocking anything, so you can compare those verdicts against your traffic and your provider’s invoice. Read the reason codes that fire on requests you know were legitimate before you enforce. Changing a policy on the day you switch it on is how you lose the users you meant to protect.

Engineer tracing a line of glowing blue nodes that step in sequence on a dark display.

Caps that survive a rotated number block

Per-number limits stop working the moment the attacker changes blocks, so set caps at every layer: per device, per number, per network and per number block. Numbers and cards are kept only as keyed hashes, so a cap can follow a device without storing the raw value. When a block burns out, the device and network caps still stand.

FAQ

Two hands at a dark desk lit by blue light while a code is about to be sent.

Is AIT the same as SMS pumping?

Yes. Artificially inflated traffic, SMS pumping and OTP toll fraud name the same pattern from different seats: messaging providers say AIT, risk teams say pumping. The mechanism is identical.

Does Prynt send the OTP?

No. Prynt returns the decision and your server sends the message through your own provider. Control is a paid plan within the Prynt platform.

Will this slow down legitimate users?

Only on mixed requests. Step-up means a captcha or an email check on the exceptions — new device, proxy, first-seen number — not on every code. The mirror period shows how many real users land in that bucket before you enforce anything.

Next step

AIT will not appear in your account dashboards, and it rarely stops on its own. Put a decision in front of the endpoint that sends the code: see how Prynt Control fits, wire it up with the docs, and read our notes on bot detection. Start in mirror mode, then enforce.

Try it free

Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.

Keep reading