Card testing prevention, short version

Card testing prevention is the work of blocking the small, fast payment attempts criminals use to check whether stolen card numbers still work. The pattern hides well. A burst of $0.50 to $2 authorizations from one IP range, a scatter of declines, then a chargeback wave six to eight weeks later. You stop it with rate limits, risk rules, and verification steps that fire before the authorization request ever leaves your checkout.

I treat it as a bot problem first and a fraud problem second. The card data is already stolen by the time it reaches you. What you control is whether your payment page becomes the free validation service.

What the attack looks like from the merchant side

Attackers run scripts against your checkout or against your payment API directly. They pull card numbers from a dump, submit each one, and keep the ones that return an approval. The volume per card is small. The volume per hour is not.

Signals I watch for:

  • Authorization attempts per minute spiking well above your normal peak, especially outside your selling hours
  • Many different card numbers hitting the same customer email, device fingerprint, or IP subnet
  • A high ratio of declines to approvals on low-value orders, then approvals drying up as the BINs get burned
  • Failed AVS and CVV checks clustered on the same shipping address or none at all
  • Disposable email domains, datacenter IPs, and mismatched billing country versus IP country
  • Checkout sessions that complete in under a few seconds, faster than a human can type a card number

None of these proves anything alone. Three of them together on one session is enough for me to challenge it.

Controls that actually reduce volume

Rate limits at the edge

Cap authorization attempts per IP, per session, and per card fingerprint. Put the limit in front of your application, not inside the payment handler, so junk traffic never reaches your processor. A sliding window of a few attempts per minute catches scripted clients without touching normal shoppers.

Require CVV and AVS, and enforce the result

Collecting the security code and running address verification is table stakes. The part people skip is acting on a mismatch. If your rule is to authorize anyway and sort it out later, you have handed attackers a working oracle. Decline on CVV failure and require a second factor on AVS mismatch for new customers.

Challenge suspicious sessions, not everyone

Invisible bot scoring plus a challenge on the risky slice works better than a CAPTCHA on every checkout. Every friction step you add to the happy path costs conversion. Every step you skip on a datacenter IP costs you a chargeback.

3-D Secure on new and high-risk traffic

Strong customer authentication shifts liability and kills a lot of testing outright, because the attacker usually cannot pass the issuer challenge. Applying it selectively keeps friction off your repeat buyers.

Block disposable identities

Disposable email domains, known proxy and VPN ranges, and repeat device fingerprints tied to prior fraud are cheap signals. Keep a rolling deny list and expire entries so you do not punish shared corporate networks forever.

Detection beats blocking on its own

Rules catch the lazy attacks. The patient ones spread attempts across weeks and IPs. That is where monitoring earns its place: track authorization attempt rate, decline rate by BIN, and the count of distinct cards per device. Set an alert on a sudden jump in attempts per session and route it to someone who can act the same day.

Share what you learn. Contributing confirmed fraud data back to your processor and to industry fraud networks makes the next merchant's blocklist better and shrinks the pool of live numbers.

When you are already under attack

  1. Raise the bar immediately: tighten velocity limits, turn on a challenge for all new sessions, and block the offending IP ranges.
  2. Contact your acquirer or payment provider the same day. They can see patterns across merchants and may already know the BIN range.
  3. Pull the last 30 days of low-value approvals and look for a shared fingerprint or email pattern.
  4. Issue refunds on orders you can identify as test transactions before they become chargebacks. A refund costs you the interchange. A chargeback costs you the fee, the product, and a mark on your ratio.

Keep your false decline rate honest

Every control you add blocks some real customers. Watch approval rates by segment for a week after each rule change. If your good-customer approval rate drops more than a point or two, the rule is too broad. Card testing prevention is a tuning job, not a switch you flip once.

The goal is simple to state. Make your checkout expensive to abuse and cheap to use. Everything above is in service of that balance.