Monitoring a Magento Store After CMS-Block Malware Cleanup
The visible loader can disappear while the access that created it remains. Sansec's Magento campaign used the cms_block table, and its newest domain appeared on 83 stores. The broader campaign reached 953 stores since July 2025. Monitoring after cleanup should cover the content layer, the identities that can write it, the files that may execute on the server, and the browser requests that leave the store.
Watch content changes
Track changes to cms_block and cms_page, including create, update, and delete events. Record the administrator or service identity, source address, timestamp, and affected identifier. Compare the change with the merchant's approved content workflow.
Alert on new script tags, external domains, encoded content, or unexpected HTML in blocks that previously contained plain text. These rules need tuning with the business. A marketing team may add a legitimate analytics script during the same week that an attacker adds a loader. The review should identify an owner and business purpose, not simply block every change.
Keep the campaign domains in a detection list. Sansec documented manulfie.click, neswnsssn.click, lanhd6549tdhse.top, additionalver.com, deepveindrilling.com, and two additional domains used on a few stores. Search both current content and historical records.
Watch identity
Monitor administrator creation, role changes, password resets, new API tokens, integration updates, and sessions from unexpected locations. Review database users and deployment credentials because the CMS content can be changed outside the admin interface.
Sansec reported a fully patched Magento 2.4.7-p10 store where a rogue administrator account was added before the loader. The patch did not answer the identity question. Monitoring should connect a new account with the content change that follows it.
Alerting on every account event will create noise. An operator can use a small set of high-value rules: a new administrator, a privilege change, a new integration token, a password reset for an inactive account, or a content change made by an identity that normally does not publish.
Watch files and execution
Monitor media and upload directories for new executable content. Review files under pub/media/custom_options/quote/ and other writable paths. Content type should be checked from the file itself. A .png extension does not prove that the file is an image.
File-integrity monitoring should cover application code, extensions, scheduled jobs, and server configuration. A clean core release does not cover every custom module. Compare the deployed code with a known build and investigate changes made outside the pipeline.
Watch the browser path
The loader fingerprints browsers and can return different content to automated clients. Sansec identified checks for navigator.webdriver, Selenium globals, ChromeDriver properties, WebGL, CPU, memory, screen size, time zone, languages, and client hints. It also found code stored in localStorage and executed later through an invisible element.
A browser monitor should record JavaScript requests, console errors, storage changes, and redirects in a controlled profile. Automated page checks can remain useful for raw HTML and CMS evidence, but they need a documented client profile and an understanding of what they cannot see.
Watch external references
ZoomEye returned 4 indexed public pages for http.body="deepveindrilling.com" on November 16, 2026. The same session returned 134,283 all-scope observations for app="Magento". The product query is a broad public baseline. The body query is a small indexed-page observation.
Neither result is a live inventory of infected stores. The Sansec research count of 83 stores on the newest domain comes from a different collection method. External queries are useful for finding public references, owned assets, and changes over time when the exact query and scope are retained.
Test the cleanup
After removing the loader and its access path, run a controlled test that compares the storefront with an approved baseline. Inspect the database, the administrators, the uploaded files, the web logs, and the browser session. Confirm the relevant Adobe fixes on the actual deployed code.
Then watch for recurrence over a period long enough to cover normal content updates and deployment activity. Pay attention to old loader domains as well as the newest one. Sansec found some stores still loading expired domains that had been re-registered by a parking service.
Build a response record
Monitoring is easier to defend when each alert resolves to a fact. A content alert should identify the block and writer. An identity alert should identify the account and permissions. A file alert should identify the hash and execution context. A browser alert should include the request and response evidence.
The campaign shows why one normal page load cannot close an incident. The useful closure record connects the visible content to the access that created it and shows that both remain under control.
Top comments (0)