PluginsAppsPricingConsultingAboutBlog (opens andrewbaker.ninja in a new tab)Help Get in touch
👁13views

← CloudScale Plugin Help/CloudScale Cyber and Devtools: Free WordPress Security, AI Penetration Testing & Developer Toolkit

Virtual Patching

When a plugin or theme you have installed has a published vulnerability and you cannot update yet, a virtual patch refuses the one kind of request that vulnerability needs and leaves everything else alone. Patches come from a signed CloudScale list, apply only to what is installed here at an affected version, and stop applying when you update. Off by default, and Log only first.

WordPress virtual patching with signed rules for known plugin vulnerabilities, in Log or Block mode
On a phone
WordPress virtual patching with signed rules for known plugin vulnerabilities, in Log or Block mode, as it appears on a mobile screen
The same panel at 390px, showing how it reflows. Cropped to the first screens.

🩹 Patch the Request, Not the Plugin

The fix for a published vulnerability is to update. When you cannot update today, because the fix breaks something or is not out yet, a virtual patch refuses the one request shape the vulnerability needs, such as a non-number in the id field of one AJAX action, and nothing else.

How It Works

  • A signed list, fetched daily. CloudScale publishes the patches as one signed list. The plugin checks the signature against a key built into it and refuses the whole list if it does not match, is older than the one it has, or is not understood.
  • Only for what you have installed. Each patch names a plugin or theme and the versions it affects. Patches for anything not installed here, or already updated, are ignored, and a patch stops applying the moment you update.
  • A description, never code. A patch names a request (an AJAX action, REST route or path, and one parameter) and one of a fixed set of checks the plugin already carries. A wrong patch can at worst refuse one kind of request to one component.
  • Drafted, tested, reviewed. Each patch is drafted from the difference between the vulnerable and fixed release, checked automatically against both, and reviewed by a separate AI agent before it is signed.

Where the Patches Come From

  • The source. Advisories come from wpvulnerability.net, the same database CVE Alerts checks your site against. Releases come from WordPress.org.
  • What qualifies. A WordPress.org plugin that sites running CloudScale use, with an advisory published in the last year, rated medium severity or above, and a fixed version released on WordPress.org.
  • Every night. CloudScale looks for new qualifying advisories each night and drafts a patch for each one. A separate AI agent reviews every draft, and only a patch it approves is signed and added to the list. CloudScale can revoke any patch.
  • Figures you can check. How many patches are live, how many were added each month, how long a patch took from the advisory’s publication to approval, and how many drafts are waiting for review are all public at api.cloudscale.consulting/v1/virtual-patches/stats. The figures come from the same records that produce the patches.

Safe to turn on

  • Off by default. Nothing is fetched until you choose a mode. Nothing about your site is sent to fetch the list.
  • Log only first. Each would-be refusal is recorded in Recent Matches as vpatch: plus the patch id. Block answers those requests with a 403.
  • Never in your way. Signed-in editors and administrators and scheduled tasks are never checked, and a firewall exception for rule vpatch on a path lifts patches there.
  • Not every vulnerability can be patched this way. A missing permission check, for example, has no bad input to refuse. Updating is still the fix; the patch buys you time.
← Back to all sections