How To Break Multi-Tenancy Again and Again ...and What We Can Learn From It
Lorin Lehawany & Sven Nobis, ERNW
Abstract
Namespace-based multi-tenancy is challenging to implement and less effective than control-plane isolation. Thus, the latter is the standard today. But is this really true? Workloads such as machine learning, pipelines, or scripting capabilities can introduce unobvious multi-tenancy in clusters and become increasingly popular. So the question is: How to isolate those workloads from each other securely? Pod Security Standards, Network Policies, and Admission Controller are well-adopted, but is it enough? The answer is no: This talk presents real-world exploits in Kubeflow, Istio, and Traefik to bypass threat boundaries between namespaces and workloads. Based on these examples, this talk presents a methodology for assessing complex environments with isolation problems and guides how to address them.
