> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ankra.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Cluster Hardening

> The network, SSH, and control plane hardening Ankra applies when it provisions clusters, and the security add-ons you can layer on top.

When Ankra provisions a cluster, it does not just create VMs and install Kubernetes - it applies a set of infrastructure defaults that reduce the cluster's attack surface from the first boot. This page describes what those defaults are.

<Note>
  This page is about clusters **Ankra provisions**. Imported clusters keep the posture you built them with - Ankra manages their workloads through the agent but does not modify their infrastructure, and they have no Ankra bastion.
</Note>

***

## Network posture

* **Tag-scoped cloud firewalls** - each node group and the bastion get their own firewall rules, scoped by provider tags rather than static IP lists, so rules follow nodes as they are added and removed.
* **Private networking with NAT** - where the provider supports it, nodes are placed on a private network with no public IPs; their egress is routed through the bastion or gateway.
* **Bastion jump hosts for SSH** - the bastion is the only host with a public IP, and SSH to nodes goes through it. See the [Hetzner architecture](/reference/hetzner#architecture) for a concrete example of this layout.

***

## SSH keys

Ankra generates a **4096-bit RSA key pair per cluster**. The key is held authoritatively in Vault and is never stored in the cluster definition - exporting, cloning, or committing a definition to Git cannot leak key material.

***

## Control plane joins

kubeadm join tokens are generated on demand when a node joins and **expire after 15 minutes**. There are no long-lived join secrets to rotate, leak, or forget about.

***

## Security add-ons

The add-on catalog covers the common security building blocks: **cert-manager**, **Kyverno**, **Trivy Operator**, **External Secrets**, **Sealed Secrets**, **Cilium**, **Istio**, and **Vault**. They install like any other [add-on](/concepts/addons) - visible in the platform, versioned, and removable - so there is no hidden security infrastructure you cannot inspect or take out.

A default security baseline for newly provisioned clusters is rolling out, so new clusters can start with a curated set of these add-ons instead of an empty slate.

***

## OS-level options at creation

The cluster create flow includes OS-level audit and hardening options. Enable them at creation and they are applied as each node is provisioned, including nodes added later through scaling.

***

## Related

<CardGroup cols={2}>
  <Card title="Security at Ankra" icon="shield-halved" href="/security/overview">
    The trust model these defaults sit inside.
  </Card>

  <Card title="Cluster Security" icon="bug" href="/guides/cluster-security">
    Vulnerability posture once the cluster is running.
  </Card>

  <Card title="SOPS Encryption" icon="lock" href="/guides/sops">
    Encrypt sensitive values in your Stacks.
  </Card>

  <Card title="Ankra Agent" icon="satellite-dish" href="/concepts/cluster-agent">
    The hardened, outbound-only agent pod.
  </Card>
</CardGroup>