Skip to main content

Security Exposure

The Jupyter Server documentation states that access to the server means access to running arbitrary code, and that connecting to kernels or terminals is equivalent to full permissions. Binding the server to all interfaces for remote use makes that code execution reachable from the network. Token authentication is on by default, and the documentation marks disabling it by setting empty token and password as not recommended unless access is handled at another layer.

Hardening

  • Keep Jupyter bound to localhost, its default, and handle remote access through an authenticated layer such as a reverse proxy.
  • Never disable token or password authentication on a server reachable from other hosts.
  • Enable TLS when the server must listen on a network interface.
  • Use JupyterHub for multi-user access instead of sharing a single-user server.

Monitoring

Alert on 8888 listeners bound to non-loopback addresses, and review server logs for logins and new kernel or terminal sessions from unexpected addresses.

Frequently Asked Questions

What runs on port 8888?→

Jupyter Server and Jupyter Notebook use 8888 by default. IANA's registration for the port is NewsEDGE (ddi-tcp-1 and ddi-udp-1).

Is it safe to expose Jupyter on port 8888?→

Only with authentication and TLS in place. Jupyter's documentation says server access means the ability to run arbitrary code.

How do I access a remote Jupyter server without opening 8888?→

Keep the server on its default localhost binding and put an authenticated proxy in front of it. For shared deployments, the Jupyter documentation names JupyterHub as the official multi-user solution.

Port assignments from the IANA Service Name and Transport Protocol Port Number Registry. Services can be configured to run on any port, so traffic on port 8888 is not guaranteed to be Jupyter. Exploited-in-the-wild data from the CISA KEV catalog (CC0).