kro (Kube Resource Orchestrator) 0.1.0 before 0.2.1 allows users (with permission to create or modify ResourceGraphDefinition resources) to supply arbitrary container images. This can lead to a confused-deputy scenario where kros controllers deploy and run attacker-controlled images, resulting in unauthenticated remote code execution on cluster nodes.
The product receives a request, message, or directive from an upstream component, but the product does not sufficiently preserve the original source of the request before forwarding the request to an external actor that is outside of the product’s control sphere. This causes the product to appear to be the source of the request, leading it to act as a proxy or other intermediary between the upstream component and the external actor.
If an attacker cannot directly contact a target, but the product has access to the target, then the attacker can send a request to the product and have it be forwarded to the target. The request would appear to be coming from the product’s system, not the attacker’s system. As a result, the attacker can bypass access controls (such as firewalls) or hide the source of malicious requests, since the requests would not be coming directly from the attacker. Since proxy functionality and message-forwarding often serve a legitimate purpose, this issue only becomes a vulnerability when: