Using the API server proxy to bypass network policies

I’ve been doing some work for an upcoming talk on Kubernetes Multi-tenancy security, and as part of that I was thinking about one of my favourite topics in Kubernetes security, SSRF. As I was doing that I realised that applying a known existing weakness in Kubernetes security could be used by attackers to get unauthorised to other tenant’s workloads in a multi-tenant cluster. In 2019 Kinvolk published a blog on abusing Kubernetes API server proxying which talked about the idea of overwriting a pod’s IP address to allow access to an external IP address with the API server’s network position, and it occurred to me that is likely to apply to internal IP addresses too. The mitigating discussed in that blog which was placing the API server in the same network position as worker nodes, made sense when the attacker was trying to access IP addresses external to the cluster, but doesn’t really apply to an attacker who’s trying to get access to an internal resource that should be blocked. Setup…