DEV Community

facilishz
facilishz

Posted on

Closing the GL.iNet Incident: An Evidence Checklist

Closing the GL.iNet Incident: An Evidence Checklist

The end of an incident is not the moment a vendor says a cloud path is blocked. It is the point when an organization can show which devices were in scope, which controls were changed, which checks passed, and which questions remain open. GL.iNet's November 20, 2026 notice gives enough information to build that record, but not enough to declare every customer device safe.

The company identified a vulnerability in its cloud remote-management services and said an unauthorized party exploited it to access and install malicious software on a subset of routers and KVM devices. The cloud vulnerability was fixed and the attack path blocked. Existing infections remain a separate cleanup problem. Stable security firmware was scheduled for November 22, and KVM beta builds were listed for affected devices.

An evidence checklist keeps those statements from turning into unsupported claims. It also gives the next shift a way to continue without rereading the whole incident.

Device identity and ownership

List the model, serial or device identity, firmware version, network location, administrator, cloud account, and business purpose for every router and KVM. For KVMs, add the connected computers and the person or team responsible for each host. For routers, add the networks and VPN paths that depend on the device.

The list should include devices that are visible in ZoomEye but missing from the asset inventory. The saved query title="GL.iNet" returned 43,748 observations, with 20,157 on port 443, 7,156 on port 80, 416 on port 8080, and zero on port 22. These observations are not affected-device counts. They are leads for ownership checks.

Mark every device with one of four states: no evidence of compromise, indicator observed, cleaned and revalidated, or unknown. Do not use "patched" as a substitute for any of them. A patched device can still be infected, and a device without the known indicator can still have an unexamined configuration.

Cloud and local access

Review the GoodCloud account that owns each binding. Confirm the account owner, authentication method, shared users, remote Web and SSH settings, and last approved change. Remove stale accounts and unnecessary remote access. The GL.iNet documentation states that from firmware v4.10, remote Web and terminal access are enabled by default once a router is bound to GoodCloud. That default makes an explicit post-incident review necessary.

Review local access as well. The router web Admin Panel defaults to ports 80 and 443, LuCI to 8080 and 8443, and SSH to 22. KVM consoles can be reached through a local browser using a domain or IP address. The team should know which of these paths are intentional and which have been blocked.

The external query and the cloud inventory should be reconciled. A device can be bound to GoodCloud without a public port. A public product page can exist without any cloud account. Neither finding should be silently dropped because it does not fit the other map.

Malware and firmware checks

Run the vendor's process check and save the output. Look for crond -i /tmp/.log in the running process list. Record the device time, firmware version, command used, and the person who collected the result. For a KVM, check the connected computer as a separate scope item.

Confirm the installed firmware against the vendor notice. The KVM beta versions listed in the notice are 1.10.2 beta or later for RM10, RM10RC, RM1, RM1V2, RM1PE, and RM4PE, and 1.8.6 beta or later for RMQ1. The stable router and KVM builds were scheduled for November 22. If the stable build was not yet available when the check was performed, record that condition instead of claiming it was installed.

If the indicator is present, treat the device as infected and follow the vendor's cleanup instructions. If the indicator is absent, document the limited meaning of the check. A negative snapshot does not prove the absence of all malware or persistence.

Credentials and network changes

Rotate the credentials that the device could use to reach other systems. This can include local administrator passwords, cloud-account access, VPN keys, API tokens, and certificate material stored on the device. Verify that old access no longer works and that the replacement path is usable by the intended owner.

Review network segmentation, firewall rules, management source ranges, DNS settings, and any remote tunnels. Confirm that IPv6 and backup paths follow the same policy as the primary interface. Monitor for the known indicator and for unexpected outbound activity during the observation period.

Closure criteria

A defensible closure record answers five questions:

  1. Which devices were in scope, and why?
  2. Which cloud and local management paths were changed?
  3. Which firmware and malware checks were completed, with what result?
  4. Which credentials and connected hosts were reviewed or rotated?
  5. Which findings remain unknown, and who owns the next check?

The checklist does not require the organization to prove that no other GL.iNet device exists on the internet. It requires a clear boundary around the assets it owns and a record of the evidence it has. That is more useful than a broad statement that the vendor fixed the cloud path.

The incident also leaves a durable lesson. Remote management should have an owner, an expiry, a tested recovery path, and an observable footprint. When those conditions are missing, the closure checklist becomes a list of discoveries rather than a confirmation that the response is complete.

References

Top comments (0)