Most store owners have a security scanner, and most trust its report. A campaign Sansec published on 24 October 2026 is a good reason to check what that trust rests on.
Sansec, a Dutch e-commerce security firm, found that one operator has planted code in 953 online stores since July 2025, most of them on Magento. The code sat inside the stores' own databases. When a security scanner visited, it was served a clean page. When a real shopper visited, it was not.
I read Sansec's report in full, because the headline number is the least useful part. The useful part is where the code hid, how it got in, and why the usual checks missed it. All three apply well beyond Magento.
What did Sansec find?
The facts, as Sansec reports them:
- 953 stores since July 2025, mostly Magento, across five attacker domains. The newest, deepveindrilling.com, was registered on 6 October 2026, and stores began loading it the next day. It appears on 83 stores, 60 of which already carried an earlier version from the same operator.
- The code lives in the database. It sits in the
cms_blocktable, which Magento uses for the static blocks it renders on pages: headers, footers, banners. Sansec's scanner found it on 148 server installs, and never in theme files or configuration. - It is a loader. The code in the block asks the attacker's server for a script, and the server decides what to send. Sansec has not captured the final payload from the newest stage. Earlier, in December 2025, the chain sent real visitors to a site imitating the Ledger crypto-wallet brand.
- It persists in the shopper's browser. The newest version stores its payload in the browser's local storage, so a visitor who received it runs it again on later page views, even if the attacker's domain is offline or blocked.
Sansec's monitoring recorded about 52,000 reports of shoppers' browsers loading one of the earlier domains alone. That is a measure of how many real visits the code reached.
How did the code get in?
Three routes appear in the report.
An unpatched upload flaw. In March 2026, the operator used PolyShell (CVE-2026-48356) to upload web shells and injected the loader into 471 stores within one hour. Web shells were found disguised as PNG images in a media folder.
Missing core patches in general. Among 169 Magento 2 installs that carried the loader and were missing at least one core patch, 60% also had files uploaded through PolyShell or SessionReaper (CVE-2025-54236). On comparable installs infected by a different skimmer, the rate was 32%.
Access that outlived the fix. One store was fully patched and still got the loader. On that store, Sansec's scanner found a rogue administrator account and, eight days later, a backdoor. Another store received the loader an hour after a clean scan.
That last point matters most. A store can apply every patch and still be compromised. A patch closes the door the attacker came in through. It does not remove an admin account or a backdoor they left behind.
Why did the scanners miss it?
Because the attacker chose who saw it. The loader fingerprints each browser before doing anything. It checks for signs of automation, such as the navigator.webdriver flag and traces left by Selenium and ChromeDriver, and collects details like the graphics card, CPU cores, screen size, time zone and language. Bots and scanners get nothing.
Sansec tested this. An automated Chrome visit from urlscan.io received a normal page with no added content.
Most website scanners work exactly this way: an automated browser in a data centre loads your pages and looks for known bad code. If the attacker's server recognises that visitor and stays quiet, the report says clean. Meanwhile the content block that does the loading is a database row, not a file, so a file-integrity check never sees it either.
Does this matter if you are not on Magento?
The specific checks below are for Magento. The pattern is not. Every modern store keeps some of its page content in a database or admin panel rather than in code files:
- content blocks, banners and page-builder sections
- "custom HTML" or "custom code" sections in a theme editor
- tracking and marketing scripts pasted into a header setting
- app or plugin settings that inject scripts onto the storefront
Each of these is a place where a script can be added by anyone with admin access, and none of them shows up in a scan of your code repository. If your store has staff accounts without two-factor authentication, an agency login nobody removed, or an old plugin that is no longer updated, you have the same exposure in a different form.
What should a Magento store check now?
Sansec's recommendations, condensed:
| Check | How |
|---|---|
| Search the database for the loader | Look in cms_block and cms_page for fpCode, se_referrer and localStorage.getItem (the domains are base64-encoded, so search for the code patterns). Check update_time for recent writes |
| Look for uploaded web shells | Search pub/media/custom_options/ and pub/media/customer_address/ for any file other than .htaccess, and search for accesson.php |
| Review admin accounts | Check admin_user for accounts you do not recognise |
| Look for unknown processes | Check the server for processes nobody can explain |
| Patch | Install APSB26-73 (PolyShell) and APSB25-88 (SessionReaper) |
| Clean up properly | Removing the script does not remove the attacker's database access. After cleanup, change the database, admin and API integration passwords |
If you have an agency or a developer on retainer, send them this table today and ask for the result in writing.
What should any online store do?
These come from the same lessons, applied to any platform.
- List every script your pages load. Open your checkout page, look at the network requests, and write down every outside domain. Anything nobody can explain is a question to answer this week.
- Monitor what shoppers' browsers actually load. A Content Security Policy in report-only mode makes each visitor's browser report scripts from domains you have not approved. This is how Sansec's monitoring saw these loads, and it is the one check the attacker cannot hide from by fingerprinting scanners.
- Treat admin access as the main risk. Two-factor authentication on every admin login, no shared logins, and remove accounts for anyone who has left, including agencies.
- Patch on a schedule, not when you remember. Most of the infected Magento installs Sansec examined were missing patches.
- Plan the clean-up, not just the removal. If you find unknown code, assume the attacker still has a way in until you have found and closed it, then rotate every credential the store holds.
CERT-In's recent advisory on AI agents made a similar point: attackers rarely need anything clever when access is left open. And the PHP2Shell vulnerabilities showed how quickly an unpatched flaw becomes a mass campaign.
Where we help
This is part of how we build and maintain Ecommerce Websites at INovaBeing: a known list of scripts on every page, admin access with two-factor authentication, patches on a schedule, and monitoring of what customers' browsers load, not only what a scanner sees.
If you are not sure what your current store loads, who has admin access or when it was last patched, a Technology Diagnostic is a sensible first step. It maps what exists before anyone changes anything.
What I take from this
A clean scan tells you that a scanner was shown a clean page. That is not the same as your customers being shown a clean page, and this campaign was built around the difference.
The fixes are not new: patch, limit admin access, know your scripts, and watch what real browsers load. What has changed is that skipping them now goes unnoticed for months. Some of these stores had carried the loader since October 2025.
Sansec's full report, with the indicators and database queries, is on its research site. It is worth forwarding to whoever maintains your store.




