Getting Started with EderaON

Containers are not a security boundary. They're kernel constructs — namespaces and cgroups — running on a shared Linux kernel. One container escape compromises every workload on the node. And with AI now capable of discovering and chaining kernel vulnerabilities through a single API call, the window for treating this as an acceptable risk is closing. In this hands-on workshop, James Petersen (Founding Engineer, Edera) walks through the problem from first principles and then demonstrates microVM-based isolation live using EderaON. Hosted by John Speed Meyers (VP of R&D, Edera), the session covers the shared kernel threat model, real container escape CVEs, and a step-by-step walkthrough of launching zones, running workloads, and deploying Kubernetes pods inside Edera's per-workload microVMs. Topics covered: Why containers are just a UX abstraction: namespaces, cgroups, and what the kernel actually sees The shared kernel blast radius: how one container escape compromises every tenant on the node Real container escape CVEs: Leaky Vessels, the NVIDIA container toolkit escape, and CopyFail How AI is accelerating vulnerability discovery — including kernel-level exploits found by Mythos Use cases for microVM isolation: multi-tenant Kubernetes, AI agent sandboxing, untrusted code execution, and Docker-in-Docker Live demo: installing Edera, launching zones, running containers inside microVMs, and deploying Kubernetes pods with a single runtime class change Dynamic resource allocation: memory ballooning and CPU overcommitment without static reservation Per-workload kernel events: scraping /proc for metrics scoped to a single process instead of the entire host Workload identity via SPIFFE/SPIRE: cryptographic identity from the hypervisor for zones and workloads 0:00 Introduction 0:48 Getting Started With EderaON 2:07 Containers Are Not a Security Boundary 5:06 The Shared Kernel Blast Radius 6:48 Container Escape CVEs: Leaky Vessels, NVIDIA Toolkit, CopyFail 8:32 AI-Discovered Vulnerabilities and the Closing Window 11:37 What Is Edera? Secure Execution for Software 13:49 Use Cases: Multi-Tenant Kubernetes, AI Agents, Untrusted Code 17:05 How MicroVMs Solve the Shared Kernel Problem 21:23 Build for Breach: Assume Compromise by Design 24:21 Live Demo: Installing Edera and Launching Zones 34:53 Running Containers Inside MicroVMs 39:29 Dynamic Resources and Per-Workload Kernel Events 42:59 Kubernetes Integration: One Runtime Class Change 49:19 How Edera Fits With AI Agent Sandboxing 52:33 EderaON Resources and Next Steps