DGX Spark/Guides/Security hardening a rented GPU instance
Security

Security hardening a rented GPU instance

By Samuel Seidel · Updated September 23, 2026

This page is a checklist for hardening a rented Spark from the inside: SSH keys, network exposure, non-root services, secrets, and what survives a stop. A rented Spark gives you root inside your container and a public network path; what runs there, and how exposed it is, is your configuration. You do not get root on the host: GPUwerk configures the machine and keeps administrative access to the host and platform infrastructure, as the DPA describes. That infrastructure layer (physical security, isolation between tenants, and so on) is outside this page, and the site does not publish a specific security certification to cite here.

SSH: the door you actually control

You don't need an SSH key to start: the console's browser terminal works without one. If you add your own key for SSH, password authentication is disabled across the fleet, per GPUwerk's SSH key guide, so your key is the only SSH credential. That already removes the most common brute-force target. From there, the hardening is on you:

Network exposure: expose only what needs exposing

Every instance gets SSH on its own hostname at the standard port 22. Model-preset instances also get a public HTTPS endpoint that requires an API key with the inference scope; base images (DGX OS, Ubuntu) get none, per the SSH guide. Anything else you bind to a port is your call, and the default should be not public.

Don't run services as root just because you land there

The browser terminal and SSH both put you in as root, and it's tempting to run everything that way since nothing stops you. Don't. A service running as root that gets compromised (through a dependency vulnerability, a malicious model artifact, or an exposed debug endpoint) gives an attacker root in your container immediately, rather than the more limited blast radius of a service account.

Secrets: don't leave them in shell history or plaintext env files

API keys, model repository tokens, and any credentials for services you call out to need the same discipline you'd use anywhere else.

Ephemeral vs persistent: pick deliberately

A stopped instance keeps its workspace and hostname exactly as they were, billed at 75% of the running rate, per GPUwerk's pricing page. A terminated instance is wiped for the next tenant and your workspace is deleted with no copy kept. That split is also a security decision, not just a billing one:

What this page is not claiming

This is in-instance hardening: what you configure once you have root inside your container on a rented machine. GPUwerk keeps administrative access to the host and platform infrastructure, so this checklist does not make you the only party with access to the machine. GPUwerk's own infrastructure security, isolation between tenants at the platform level, and physical access controls are a separate posture that this page does not describe or vouch for beyond what is already published on GPUwerk's own pages. No specific security certification is claimed here; if that matters for your compliance requirements, ask directly rather than infer one from this guide.

A first engagement can help scope a hardening checklist against your specific compliance requirements before you move real workloads onto a rented Spark.

Start from a clean instance and configure it yourself.

Root in your container, a browser terminal that needs no SSH key (key-only SSH if you add one), and a workspace that persists until you terminate. A dedicated Spark from $0.79/h, billed by the minute. First top-up: pay $20, get $30 in credit.

Create account Read the SSH guide