Red Hat has disclosed CVE-2026-66794, an important-severity server-side request forgery (SSRF) vulnerability in the cluster-proxy-addon component of the Multicluster Engine for Kubernetes.
This flaw has a CVSS v3.1 score of 9.3. It could allow an unauthenticated remote attacker to use a publicly accessible route to access otherwise isolated services across managed clusters.
Published on August 19, 2026, the vulnerability arises from insufficient authentication and authorization protections on the add-on’s user-facing route. An attacker with access to this route can manipulate URL path segments, causing the cluster proxy to forward requests to arbitrary services on connected managed clusters.
The vulnerable component serves as a proxy between user-facing endpoints and services within managed Kubernetes clusters. Red Hat has stated that the affected route can be exploited without requiring credentials, allowing an attacker to bypass typical access controls and relay requests through the proxy.
This turns the exposed proxy into an SSRF vector: instead of directly reaching an internal endpoint, the attacker instructs the vulnerable service to send the request on their behalf.
Because the request originates from infrastructure trusted by the managed cluster environment, network segmentation and firewall rules may no longer provide the expected level of protection.
For instance, an attacker could craft a request with manipulated path segments that make the proxy contact an internal dashboard, API endpoint, or service that is not publicly accessible.
The exact outcome will depend on the available services, their authentication requirements, and the configuration of network policies.
Red Hat identifies the following components as affected: multicluster-engine/cluster-proxy-addon-rhel9 and multicluster-engine/cluster-proxy-rhel9.
At the time of disclosure, the advisory did not include security errata for either component. Red Hat advises that unless explicitly marked as unaffected, earlier packages within applicable minor-version streams should be considered vulnerable.
The CVSS vector is CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N, indicating network reachability, low attack complexity, no required privileges, and no user interaction.
The primary confirmed risk of this vulnerability is a high confidentiality impact; the integrity impact is rated low, and the availability impact is rated as none.
Initially, Red Hat assessed the potential impact as higher. However, it revised the score after confirming it as an unauthenticated SSRF rather than a remote code execution vulnerability. Consequently, the vendor classified it as Important rather than Critical in its severity framework, despite the high numerical score.
Organizations using the Multicluster Engine should immediately restrict access to the cluster-proxy-addon user-facing route. Red Hat recommends implementing firewall controls to ensure that only trusted networks and authorized source addresses can access the exposed port of the route.
Security teams should also:
CVE-2026-66794 is classified under CWE-918, a weakness category associated with proxying requests to unexpected hosts, ports, or protocols. In Kubernetes management environments, this risk is heightened because a single exposed proxy can connect to multiple managed clusters.
Prevent incidents due to slow investigations. Power your Tier 1 with threat intelligence from 15K SOCs: Integrate TI Lookup in your SOC
OpenAI has temporarily slowed the development of its latest frontier AI models after initial testing…
ToxicPanda 2.0, an evolved Android banking Trojan that significantly expands its fraud, device control, and…
Cisco has issued security updates for a high-severity vulnerability in Cisco BroadWorks that could allow…
Threat actors are pairing fake CAPTCHA verification pages with a commercial malware loader capable of…
Splunk has released a security hardening update addressing 17 vulnerabilities across several applications and add-ons,…
Threat actors are increasingly abusing Microsoft 365 identity sessions rather than deploying malware, as shown…