No vendor advisory, no incident report and no researcher write-up is circulating publicly under the name “Convertr skimmer attack”at the time of writing. If you have been asked about one by a director, an acquirer or a customer, the first useful thing to establish is that the name in the question describes a suspected delivery route, not a confirmed event and not an attacker. The five checks below settle the question in the order that actually reduces your exposure, and they work for any supplier-named skimmer scare, whichever vendor is in the headline next month.
What a “Convertr skimmer attack”claim actually tells you
Start by separating two things that get welded together in a panic. A supplier name tells you which file somebody suspects. It does not tell you who wrote the code, and it does not tell you that your store loads that file anywhere near a card number.
The Magecart label makes this worse. It is a name for a method rather than a group, coined from Magento plus shopping cart when researchers first catalogued the technique, and now used for dozens of unrelated crews who share nothing but an approach. Asking “which group is this”burns days and changes nothing. Asking “does that origin execute in the context where cards are typed”changes everything.
Every claim of this shape resolves into one of three routes in. A compromised third-party supplier, serving a modified file from its own CDN. Injected first-party code, served from your own domain. Or an unpatched platform or extension, where the attacker never touched a supplier at all. Check 1 through Check 4 test the first. Check 5 tests the other two, and it is the one most teams skip.
What to do in the first hour
Resist the urge to rip the tag out mid-trade. Lead-capture and marketing scripts are often wired into order confirmation, fraud screening handoffs or consent logic, and pulling one at 11am on a Tuesday can produce an outage that looks like a breach response gone wrong. Removing it also destroys the evidence you need.
Instead, capture state. Load the payment step in a browser with the network panel recording, save a HAR file, screenshot the script origins, note the date and time, and download the live copy of the suspect file to disk. Then work through the checks. If Check 1 comes back clean, you have answered the director’s question in under an hour with evidence attached.
Check 1: does the vendor script load anywhere near your card fields?
Open the actual payment step, not the basket and not the homepage. Record every script origin the browser requests, including the ones requested by other scripts. A loader pulls children, and children pull children.
Then go to the tag manager container, because that is where the answer usually lives. Marketing and lead-capture tags are almost never added to a checkout template on purpose. They are published site-wide, on an all-pages trigger, by someone in marketing who was never asked whether the payment step counted as “all pages”. A merchant can say honestly that the vendor is not in the checkout code and still be loading it on the card-entry page. Auditing the tag manager container rather than the template is the difference between a true answer and a confident one.
Hold the distinction between two findings:
- Loads on the site. A fact about your marketing stack. Worth knowing, not an exposure.
- Loads in the card-entry context. A fact about your risk. The script can read the DOM of the page where card data is typed, or draw over it.
If the origin does not appear on the payment step at all, write that up with the HAR file, the screenshot and the date, and close the question. Do not reopen it on a hunch. Reopen it when the evidence changes.
Check 2: read the code the browser is really executing
Version numbers are claims. The file served to your customers is the only artefact that matters, so fetch the live copy from the vendor CDN and diff it against a known-good copy, an archived build or yesterday’s download. A diff with a timestamp is worth more than any supplier reassurance.
What you are looking for, in rough order of how often it appears:
- Event listeners bound to input, change, keydown or form submit handlers, on fields the vendor has no business reading.
- A MutationObserver waiting for a card form to appear, which is how a skimmer survives a single-page checkout that renders the payment step late.
- Base64 or hex blobs, long string arrays, or a decoder function assembling a domain name at runtime.
- A beacon to a lookalike host: the brand name with a hyphen, an extra word, a different top-level domain, often fired on sendBeacon or as an image request so it survives page unload.
Look for the second stage as well. Skimmers are routinely delivered by a short, clean, boring loader that fetches the real payload only when the URL contains something like “checkout”, “onepage”or “payment”, and serves nothing at all to anyone else. That pattern is designed for exactly the review you are doing. Test the file with a referrer and path that match your payment step, not a bare curl against the CDN.
One caution. Obfuscation is a question, not a verdict. Plenty of legitimate vendor bundles are minified past readability, mangled by a build tool and shipped with inline sourcemaps stripped. Unreadable does not mean malicious. Unexplained network destinations do mean something, and so does a file that behaves differently depending on the path. If you want the longer method, the sequence for detecting malicious JavaScript on an ecommerce site goes further than a single diff.
Check 3: where card details are typed, and what that protects
If the primary account number, expiry and CVV are typed into the payment provider’s hosted fields or iframe, a compromised marketing script on your page cannot read them. The browser same-origin policy stops it. That is the single cheapest structural control available, and it is why hosted payment fields remove most skimming risk rather than merely reducing it.
Two things the iframe does not do.
It does not cover the fields outside it. Name, email, billing address, phone and often date of birth sit in your own form, in your own DOM, readable by any script on the page. That set is enough for account takeover and for convincing social engineering, and it is reportable under UK GDPR regardless of whether a card number moved.
It also does not police what is painted on top. A skimmer can render a plausible card form over the real one, collect the details, then quietly hand the shopper back to the genuine iframe for a second attempt that succeeds. The customer sees a form that “glitched once”. Confirm the frame is genuinely your provider’s: inspect the iframe src origin, check it against the provider’s documented domains, and look for any wrapper element layered above it with a higher z-index. The iframe defines the boundary. It does not guard the page around it.
Hosted fields do not exempt you from answering the script question either. Since the SAQ A changes landed, merchants who previously treated the iframe as the end of the conversation now have scripts to inventory like everyone else.
Check 4: the signals that come from outside your own logs
A working skimmer gives you no reason to look. Payments succeed, orders ship, nobody complains. So the strongest evidence usually arrives from outside the building.
Ask your acquirer for common point of purchase analysis. When issuers see fraud on a set of cards whose only shared history is your checkout, they raise a CPP alert, and a cluster of reports from customers who bought in the same two-week window outranks any amount of clean code review. Clean code today says nothing about the file that was served three weeks ago.
Read support tickets with different eyes. Reports of a double charge, an unexpected second request for card details, a payment page that “looked a bit different”, or a card declined once then accepted immediately after are all consistent with an overlay. They are also consistent with a flaky payment gateway, which is why these are signals rather than findings.
If you have Content Security Policy reporting running, check the violation feed before anything else. A spike in blocked connect-src destinations, especially to a host nobody recognises, is the closest thing to a tripwire you will get. Merchants with no reporting endpoint have nothing to look at, which is the argument for the standing control set described below.
Then put the supplier on the record. Vague questions get vague answers, so ask for specifics in writing: which exact file paths your account loads, whether the served content of those files changed between two named dates, whether the CDN origin or any build pipeline was altered, and whether the vendor has conducted or commissioned any investigation. A written answer is also what your acquirer will want to see.
Check 5: rule out your own platform and admin first
This check exists because of British Airways. The code that skimmed roughly 380,000 card transactions between August and September 2018 sat in a script served from the airline’s own domain, sending data to a lookalike host. No supplier audit would have found it. The ICO’s penalty, reduced to £20m in October 2020, remains one of the larger UK enforcement actions on record, and the route in was first-party code.
So before you escalate with the vendor, check your own ground:
- First-party files. Diff your own payment-page JavaScript against source control. Any file modified outside a deployment is the finding.
- Platform and extension patch level. In September 2020, Sansec reported that close to 2,000 Magento 1 stores were compromised over a single weekend through the platform itself. Running software past end of support is the route that needs no supplier and no insider.
- Admin entry points. Dormant accounts, shared logins, missing two-factor authentication, and any role that still permits template or file editing in the CMS. Pull the audit log for the period in question and read it line by line.
- Everything payment-adjacent. Staging environments on public URLs, legacy landing pages, and campaign microsites that collect personal data. A skimmer on a page that never sees a card still harvests enough to be reportable.
Ticketmaster UK is the mirror image, and worth holding in mind. The 2018 skimmer arrived through Inbenta, a customer support chatbot supplier that nobody had classed as payment infrastructure. The ICO fined Ticketmaster UK £1.25m in November 2020. A chat widget has no business on the page where card numbers are typed, and neither does a lead-capture tag.
Turning the scare into a standing control set
The reason this exercise takes a week is almost always that nobody knows what the payment page loads. Fix that once and the next scare is a lookup.
Run Content Security Policy in report-only mode first, for a fortnight of real traffic, and let the violation reports build the inventory for you. A mid-sized UK checkout commonly turns out to load well over a dozen distinct script origins once tag manager children are counted, and the list in the report rarely matches the list anyone would have written from memory. Report-only cannot break a payment, which is the whole point of starting there. The working CSP setup for a checkout page covers the enforcement step once the data is in.
Subresource Integrity is the one control that directly neutralises a compromised supplier file: the browser refuses to execute a script whose hash does not match. The trade-off is real and worth stating plainly. Marketing vendors change code without notice, and a stale hash blocks the file outright. The pattern that survives contact with a live store is pinning only the scripts on the payment page, alerting on integrity failures so a block is noticed in minutes rather than at month end, and leaving unpinned loading to pages where no card or personal data is entered.
Then there is the compliance floor. PCI DSS 4.0 requirements 6.4.3 and 11.6.1 have been mandatory since 31 March 2025, and between them they ask for an inventory of every script on the payment page with written justification and authorisation, plus a mechanism that detects unauthorised change to those scripts and to the HTTP headers. The text is on the PCI Security Standards Council site. Read practically, what 6.4.3 and 11.6.1 ask of your payment page is exactly the artefact that answers “are we affected”in five minutes.
Finish with a decision rule, written down, before you need it. Who runs Check 1. Who has CDN and source-control access to run the diff. Who contacts the acquirer. And who is authorised to remove a tag from the payment page, because that decision has revenue consequences and should not be made by whoever happens to be on call.
The short version
A supplier name in a skimmer story names a suspected route, not an attacker. Before anything else, establish whether that origin executes in the card-entry context, and check the tag manager container as well as the template. Diff the live file rather than trusting a version number, and expect second-stage loaders that only fire on checkout paths. Hosted payment fields protect the PAN, not the rest of your form and not the page drawn around the iframe. Clear your own platform, first-party code and admin accounts before you escalate, because British Airways and the September 2020 Magento 1 weekend both say the route is often closer to home. And if a report of a Convertr skimmer attack, or any other named-vendor scare, cost you a week of reconstruction, build the script inventory and change detection now. The next one should take an afternoon.
Frequently Asked Questions
Is there a confirmed Convertr skimmer attack?
No vendor advisory, public incident report or researcher write-up is circulating under that name at the time of writing. If the claim reached you from a headline, a forum post or an internal email, ask for the primary source before acting on it. Then run the checks against your own payment page regardless, because the answer is useful whether or not the story holds up.
How do I tell whether a third-party script loads on my payment page?
Open the payment step in a browser with the network panel recording, filter to scripts, and save a HAR file so you have dated evidence. Then open your tag manager container and look at the triggers, because site-wide tags appear on checkout without anyone adding them to the template. For a full picture, a fortnight of CSP report-only data lists origins that a single page load can miss.
Does using hosted payment fields mean a supplier skimmer cannot reach my card data?
Card number, expiry and CVV typed inside the provider’s iframe are out of reach of scripts on your page, and that removes the largest part of the risk. Name, email, address and phone in your own form are not protected at all. A script can also overlay a fake card form on top of the real frame, so confirm the iframe origin matches your provider’s documented domains and that nothing is layered above it.
Should I remove a suspect script from checkout straight away?
Capture evidence first: a HAR file, screenshots of the script origins and a saved copy of the live file, all timestamped. Removing a tag can break order confirmation, fraud screening or consent logic, and it destroys the artefact an investigation needs. If Check 2 produces an unexplained network destination or a path-gated second stage, remove it then, with a named owner signing off.
What do PCI DSS 6.4.3 and 11.6.1 require me to do about third-party scripts?
Requirement 6.4.3 asks you to inventory every script executing in the consumer browser on the payment page, confirm each one is authorised, and record a written justification for its presence. Requirement 11.6.1 asks for a mechanism that detects unauthorised change to those scripts and to the page’s HTTP headers, with alerting. Both have been mandatory since 31 March 2025, including for merchants using SAQ A, so check the current standard text and your acquirer’s expectations.
