SMSMC Blog

A certificate authority was compromised. Your containers did not get the memo

In October 2026, eMudhra CA keys were found compromised. Google, Ubuntu and Istio shipped fixes, but container images keep their own trust stores. What to check now.

By Jim Venuto

What happened in October

On October 20, 2026, Ubuntu reported that the private keys of the eMudhra certificate authority had been demonstrated to be compromised. The Istio project put it more precisely: the private keys of eight eMudhra intermediate CA certificates, including their cross-signs by Sectigo, were disclosed as compromised that day.

A compromised CA key means someone other than the CA can issue certificates that software will trust. Anyone in a position to intercept traffic can then impersonate a website or an API to any client that still trusts those certificates. Google’s bulletin for Google Kubernetes Engine, GCP-2026-071 (rated High), says unpatched clients and container runtimes that trust these authorities “are vulnerable to active network Man-in-the-Middle (MitM) interception and traffic spoofing on external TLS connections.”

The fixes came quickly:

  • Ubuntu removed the two roots that cross-signed eMudhra, emSign Root TLS CA – G3 and Comodo AAA Certificate Services, from its trusted root stores in updated ca-certificates and libnss3 packages, and published steps for Java trust stores.
  • Istio released versions 1.31.3, 1.30.7 and 1.29.10 on October 21 to distrust the compromised intermediates, and noted that Envoy, Go and ztunnel do not enforce Certificate Transparency, the check that protected browsers.
  • Google Cloud published bulletins from October 23 through November 3 covering GKE, Cloud Service Mesh and Apigee, with patched node versions and steps for customers.

The part that is not patched for you

The most useful sentence in Google’s GKE bulletin is this one: “Customer application containers do not inherit the host node’s root CA store and must be updated independently.”

Most container images carry their own copy of the list of trusted certificate authorities, baked in when the image was built. Patching the servers, the nodes, or even the managed platform does not change that copy. Google’s instruction is to “rebuild your container images using patched base images” and redeploy.

The same is true outside Google Cloud. An application in a container on AWS, Azure, a vendor’s platform or a server in the office trusts whatever was in its base image on the day it was built. Some managed services update automatically; Google notes, for example, that managed Cloud Service Mesh needs no customer action. Your own applications do not.

Why this matters to a small organization

Most SMBs and public bodies do not run Kubernetes clusters themselves. But many run containers without thinking of them that way: a vendor’s application delivered as a container, a website on a container platform, a small internal tool a contractor built. Nobody owns the question “when was this image last rebuilt?”

This incident is a clean test of whether your organization can answer three questions quickly: which applications run in containers, who rebuilds them, and how long it takes.

Pointers for this week

  1. Patch the operating systems. Apply your distribution’s ca-certificates update on servers and workstations, and refresh Java trust stores where your vendor’s guidance says so.
  2. List your container images and their owners. Include vendor-supplied images. For each, ask when it was last rebuilt from a patched base image.
  3. Rebuild and redeploy what you own; ask vendors about the rest. A short written question to each vendor, asking whether their images distrust the compromised eMudhra certificates and when they were rebuilt, is reasonable and easy to answer.
  4. Check custom trust stores. Google’s Apigee bulletin tells customers to review custom trust stores. Any application that ships its own list of trusted certificates needs the same review.
  5. Fix the process, not only this incident. Images that are rebuilt on a schedule from current base images pick up fixes like this one without a special effort.

Who does what

Leadership asks for the answer to the three questions above and decides what “rebuilt within a week” means for critical applications. Your IT team, MSP or developers patch servers, rebuild images and confirm the deployments. Vendors confirm the status of the images they supply, in writing.

Where outside help fits

A certificate authority compromise is a supply chain event: the fix is easy for each system, and the hard part is knowing every system that needs it. CyberIntelPro leads that work: the inventory of applications and images, the rebuild expectations your developers and vendors should meet, and the change list for your team. Security Medic’s vendor and third-party risk work turns the vendor question into a standing contract term rather than a one-time email.

Want to talk it through? Book a free 30-minute call with Security Medic Consulting.

Sources