Skip to content

Authenticated remote code execution via TailscalePing

Moderate
louislam published GHSA-hfxh-rjv7-2369 Nov 24, 2023

Package

No package listed

Affected versions

>= 1.23.0 && <= 1.23.6

Patched versions

1.23.7

Description

Summary

The runTailscalePing method of the TailscalePing class injects the hostname parameter inside a shell command, leading to a command injection and the possibility to run arbitrary commands on the server.

Details

When adding a new monitor on Uptime Kuma, we can select the "Tailscale Ping" type. Then we can add a hostname and insert a command injection payload into it. The front-end application requires that the field follow a specific pattern, this validation only happens on the front-end and can be removed by removing the attribute pattern on the input element.

let cmd = `tailscale ping ${hostname}`;
log.debug("Tailscale", cmd);
return new Promise((resolve, reject) => {
let timeout = interval * 1000 * 0.8;
exec(cmd, { timeout: timeout }, (error, stdout, stderr) => {

We can finally add the new monitor and observe that our command is being executed.

NOTE: When using Uptime Kuma inside a container, the "TailScale Ping" type is not visible. We can fake this information by intercepting WebSocket messages and set the isContainer option to false.

PoC

  • Authenticate.
  • Create a new monitor.
  • Select the TailScale Ping type (if not visible, see the note in the details section).
  • Insert the command injection payload inside the hostname field. (for example $(id >&2))
  • Remove the pattern requirement on the field.
  • Save and start the monitor.

image

Impact

An authenticated user can execute arbitrary command on the server running Uptime Kuma.

Remediation

There are other command execution in the codebase, they use a method spawn from the child_process module which does not interpret the command as a shell command, the same thing should be done here.

NOTE: The Tailscale CLI seems to support the -- sequence. It should be used between the ping subcommand and the hostname argument to avoid argument injection.

Severity

Moderate

CVE ID

No known CVE

Weaknesses

Credits