Card not present fraud is any payment where the card is never physically handed over: online, over the phone, mail order. For a UK merchant that label covers two completely different events, and almost every guide to card not present fraud in the UK deals with only one of them. The first is someone else’s stolen card being used on your site. The second is your checkout page being the place that card was stolen from. Same category, opposite ends.
Nothing on your server needs to break for the second one to happen. No database is dumped, no order data is exported, no ransom note arrives. A few hundred bytes of JavaScript read the card fields as a real customer types into them and post a copy somewhere else. The order completes. The money clears. Weeks later your acquirer sends you a common point of purchase report and you start working backwards.
This piece covers the table stakes quickly, then spends its weight on the supply side: where the card data used against other merchants is harvested, the three routes a skimmer reaches a payment page, and the controls that cut it off at source while satisfying PCI DSS 4.0. It does not cover fraud filter tuning, device fingerprinting or chargeback representment. Those are solved problems with vendors attached.
What card not present fraud means for a UK merchant today
Since 14 March 2022, when the FCA’s enforcement deadline for Strong Customer Authentication took effect for UK ecommerce, most online card payments have to be authenticated under the Payment Services Regulations 2017. In practice that means 3-D Secure 2: a challenge, a biometric prompt in a banking app, or a frictionless pass based on risk data shared with the issuer.
SCA works. It made plain card-on-file abuse harder. It also left gaps by design, because regulators did not want every small everyday transaction to interrupt itself. The exemptions include low-value payments (subject to cumulative limits under the onshored SCA-RTS), transaction risk analysis by the acquirer, trusted beneficiaries the cardholder has whitelisted, recurring payments and merchant-initiated transactions. Each one is a lane where a full card record can still be spent without a challenge.
Liability follows authentication. Where 3-D Secure 2 is applied and the issuer authenticates the payer, fraud chargeback liability generally sits with the issuer. Where you or your acquirer claim an exemption, it generally stays with you. That is the clean version. The messy version is first-party misuse, where the real cardholder disputes a real purchase, which arrives in your chargeback reports under the same reason codes and inflates whatever number you are using to judge your exposure.
Here is the part the generic guides skip. Authenticating the payer does nothing about how the card details were obtained. A skimmer on a payment page captures the PAN, expiry, CVV, cardholder name and billing address in a single pass, and a record that complete is exactly what gets spent against merchants applying exemptions. Protection at the point of capture and authentication at the point of spend are separate problems. Most CNP guides only discuss the second one.
Where the card data actually comes from before the fraud happens
Four main sources feed the card-not-present market: phishing pages that imitate a checkout or a bank, credential stuffing against accounts with saved cards, old breach dumps, and client-side skimming of live payment pages. The first three degrade. Phished data is patchy, stuffed accounts rarely expose full PANs, and dumps go stale as cards are reissued.
Skimmed data does not degrade, because it is harvested in real time. A JavaScript skimmer on a checkout page attaches listeners to the card fields, or reads the form on submit, or watches keystrokes, then posts the values to a collection endpoint over an ordinary HTTPS request while the payment itself proceeds normally. The customer gets an order confirmation. So does the thief.
That record arrives fresh, complete and matched. Billing address and name together mean it clears AVS. A valid CVV means it clears most basic velocity and data-quality checks. Which is why skimmed cards are priced above dump data, and why your checkout is a more attractive target than your order database.
The history is well documented. Ticketmaster UK, June 2018: the skimmer arrived through Inbenta, a third-party customer support script loaded on the payment flow, and the ICO fined Ticketmaster UK £1.25 million in November 2020. British Airways, August to September 2018: modified first-party JavaScript in the payment journey, and a final ICO penalty of £20 million issued in October 2020. Magento 1 reached end of life in June 2020, and Sansec reported in September 2020 that roughly 2,000 Magento 1 stores were compromised over a single weekend, which remains the clearest illustration of what an unmaintained platform costs. Those three cases cover the three routes in.
The three routes a skimmer reaches your checkout
A compromised third-party supplier. Chat widgets, analytics, personalisation engines, A/B testing, review platforms, tag containers. Each one loads with full page privileges, which means each one can read your card fields. Their security posture becomes yours the moment you add the script tag. Ticketmaster did not get breached; Inbenta did.
Injected first-party code. Someone gets access and writes the skimmer into something you own: a theme template, a CMS content block in the database, a payment module, an extension with a known vulnerability. There is no unfamiliar host in the network waterfall and no new script tag, because the code is inline in a file you already trust. British Airways sat in this category.
An unpatched or end-of-life platform. Magento 2 remote code execution chains get weaponised within days of disclosure. Extensions get abandoned by their authors without any announcement. A site that takes orders every day feels maintained. A working site is not a maintained site, and zero-day skimmer campaigns against Magento rely on exactly that assumption.
Then there is the blind spot. Google Tag Manager is the route that neither Content Security Policy nor Subresource Integrity defends. A nonce-based policy that permits the GTM loader effectively trusts anything a publishing user ships afterwards, and SRI hashes cannot pin a container that is designed to change. If three agencies and two former employees still hold publish rights on your container, that is your real script-deployment pipeline. Auditing publish permissions and container version history is the only control that bites there.
How to tell whether your site is feeding card not present fraud
Start outside. The first honest signal most merchants get is a common point of purchase report from their acquirer or a scheme: issuers triangulate a set of fraudulently used cards and find your store is the shared merchant. It usually arrives weeks after the compromise started. Ask your acquirer for the CPP window, the exact date range they used, because that defines the order set you need to review and the customers you may have to contact. A cluster of customers independently reporting fraud after shopping with you is the same signal, arriving less formally.
Then look inside. You are looking for outbound connections from the payment page to hosts that are not in your documented inventory, scripts that changed without a release, base64 or hex blobs in template files or CMS content blocks, and event listeners attached to card input fields. The detection work is covered in more detail in this walkthrough of checking a checkout for skimming.
The cheapest check is a synthetic order. Open a clean browser profile with no extensions, from an IP that is not your office, record the network tab, place a low-value test order with a real test card, and compare every outbound host in the log against your script inventory. Skimmers routinely suppress themselves for repeat visitors, logged-in admins and known IP ranges, which is why the clean profile and the outside connection matter.
An honest limit: CSP violation reports catch exfiltration to destinations you have not allowed. If the skimmer posts to a domain already sitting in your connect-src allowlist, such as a compromised analytics or chat supplier, no violation is ever raised. That is why a script inventory with change detection has to sit alongside CSP rather than behind it.
Cutting the harvesting at source
Keep the card out of your DOM. Hosted payment fields or a full redirect means the PAN and CVV are typed into the PSP’s iframe or page, and your scripts cannot read them, even if one of them is hostile. The trade-off is real: less design control, iframe styling limits, and conversion effects you should measure rather than assume. Take it anyway unless you have a strong reason not to.
Content Security Policy, report-only first. Run it in report-only mode for a fortnight, collect what breaks, then enforce. Pay attention to connect-src and form-action, not just script-src, because exfiltration is an outbound request and a redirected form post is a complete bypass of your JavaScript controls. There is a working checkout policy setup worth copying rather than inventing from scratch.
A script allowlist that doubles as your PCI inventory. Every script on a payment page needs to be authorised, inventoried and justified. Subresource Integrity pins third-party files by hash and genuinely defeats a compromised supplier, but the maintenance cost is not zero: when the supplier rotates their script, the hash breaks and the file stops loading. On a checkout, a silent load failure can break payment. SRI needs a monitored failure path and a named owner, not a one-off deployment.
Change detection with alerting. PCI DSS 4.0 requirement 6.4.3 asks you to manage and justify payment page scripts, and 11.6.1 asks you to detect and alert on unauthorised changes to payment page HTTP headers and script content. Both became mandatory on 31 March 2025, as published by the PCI Security Standards Council. If you self-assess under SAQ A, these still apply, and your acquirer will want evidence: the inventory itself, the justification per script, and alert records showing the detection is live. The mapping is set out in this breakdown of 6.4.3 and 11.6.1.
One note on blocklists. Skimmer families that host their payload in blockchain smart contracts, the technique Guardio Labs named EtherHiding in October 2023, cannot be taken down by serving notice on a registrar, because there is no registrar. Blocklist-based defence ages badly against that. Allowlists, inventory and keeping card fields out of your own DOM do not.
If you find a skimmer
Snapshot before you clean. A full file and database copy, plus web server and access logs, taken before anyone edits anything. Record the exact payload, its location and the exfiltration endpoint. Deleting the code first and asking questions later destroys the only evidence of how long it ran, which is the number that determines your customer notification scope.
Then find every copy. Files, CMS content blocks in the database, tag container versions, cron jobs, scheduled tasks, and admin accounts created during the window. One missed copy reinstalls the rest, and re-infection within days is the normal outcome of a partial clean. Rotate admin credentials, API keys and PSP keys afterwards.
On duties: if personal data has been compromised and the risk threshold is met, the ICO must be notified within 72 hours of becoming aware, and affected individuals must be told where there is a high risk to them. The ICO’s personal data breach guidance sets out what the report needs to contain. Tell your acquirer and PSP early, not after the clean-up, because they may need to flag the affected order set to the schemes. Then review every order in the exposure window and tell those customers plainly what was exposed and that they should watch their statements.
The short version
Card not present fraud in the UK is two problems sharing a name. Stolen cards arriving at your checkout are a filtering and authentication problem, largely addressed by 3-D Secure 2 and SCA. Cards stolen from your checkout are a client-side integrity problem, and SCA does nothing about it. If your chargeback rate is climbing and your CPP reports have started arriving, assume the supply side until you have evidence otherwise. For the attack patterns behind all of this, read next on how card details are stolen at the moment of payment.
Frequently Asked Questions
What counts as card not present fraud in the UK?
Any fraudulent transaction where the card was not physically presented: online purchases, telephone orders and mail order. UK Finance reports it as a distinct category from counterfeit and lost or stolen card fraud. For a merchant it covers both fraudulent orders placed on your site and your customers’ card data being spent elsewhere.
Does 3-D Secure 2 stop card not present fraud?
It reduces it, and it shifts chargeback liability to the issuer when authentication is applied and passes. It does not stop the data being harvested in the first place, and it does not apply where an SCA exemption is claimed, such as low-value payments, transaction risk analysis or trusted beneficiaries. Full card records captured by a skimmer are spent in exactly those lanes.
Who pays for a card not present chargeback, the merchant or the bank?
Broadly, where 3-D Secure 2 authentication was applied and succeeded, fraud liability sits with the issuer. Where you or your acquirer claimed an exemption, it generally stays with you. First-party misuse, where the genuine cardholder disputes a genuine purchase, is treated differently again and is not covered by the authentication liability shift. Check your specific acquirer agreement, as terms vary.
How would I know if my checkout is the source of stolen card data?
The usual first signal comes from outside: a common point of purchase report from your acquirer, or several customers reporting fraud after shopping with you. Inside, look for outbound connections from the payment page to hosts that are not in your script inventory, unexplained script changes and listeners on card fields. A recorded test order from a clean browser profile on a non-office IP is the quickest check.
Do I have to report a skimmed checkout to the ICO?
If personal data was compromised and the breach is likely to result in a risk to people’s rights and freedoms, you must notify the ICO within 72 hours of becoming aware. Card data with names and billing addresses will usually meet that threshold. Where the risk is high, you must also tell the affected customers. Notify your acquirer and PSP at the same time.
