Why Kubernetes Multi-Tenancy Keeps Failing: An Offensive Security Perspective

If you're running multiple teams, customers, or workloads on a shared Kubernetes cluster, you're doing multi-tenancy — whether you planned for it or not. The question is whether the boundaries you're relying on actually hold up under attack.
In this session, Lewis Denham-Parry (Staff Solutions Engineer, Edera) sits down with Iain Smart (Principal Consultant, Amberwolf), an offensive security specialist with a decade of penetration testing experience, the last eight years focused on Kubernetes and cloud-native environments. The conversation walks through real attack chains Iain has used to compromise multi-tenant clusters, why common Kubernetes primitives fail as security boundaries, and what organizations get wrong about isolation.
Topics covered:

What multi-tenancy actually means in Kubernetes: application-level, namespace-level, and direct API access models
Why taints, tolerations, and Kubernetes namespaces are not security boundaries — and the specific mechanisms attackers use to bypass them
How overly permissive RBAC and AWS metadata service access enable lateral movement across tenant boundaries
A real attack: using namespace-scoped wildcard RBAC to modify Pod Security Admission labels and escalate to privileged containers
The difference between virtualization and isolation, and why containers give you one but not the other
Where AI agent sandboxing fits into the multi-tenancy problem
Zero trust in Kubernetes: why it's a mindset, not a product, and where the model breaks down
Compliance and auditing challenges when your security boundaries don't map to what auditors expect

The session closes with audience Q&A on threat modeling adoption, and a detailed walkthrough of why namespaces aren't enough even with Pod Security Admission enforced.

0:00 Introduction
2:02 Meet Iain Smart: Offensive Security and Kubernetes Pen Testing
6:02 What Is Multi-Tenancy in Kubernetes?
10:01 Hard vs. Soft Multi-Tenancy
13:37 Why Taints, Tolerations, and Namespaces Fail as Security Boundaries
21:12 AI Agents and the Case for Secure Sandboxing
23:18 Virtualization vs. Isolation: Why Containers Give You One But Not the Other
29:12 RBAC, Lateral Movement, and the AWS Metadata Attack
41:56 Zero Trust in Kubernetes: Mindset vs. Product
46:26 Compliance and Auditing in Multi-Tenant Environments
53:19 Q&A: Threat Modeling and the Namespace RBAC Bypass

Видео Why Kubernetes Multi-Tenancy Keeps Failing: An Offensive Security Perspective канала Edera