Researcher finds Tailscale SSH flaws could expose servers to root access

A security researcher has demonstrated that a single compromised device on a Tailscale network could give an attacker root access to other reachable servers, provided two conditions are met: Tailscale SSH is configured to allow port forwarding and the attacker already controls one device that the network’s access rules permit to connect to the target.

The findings, published on 25 October by Erica Windisch of Cyberdione Labs in a report named Tailworm, tested Tailscale version 1.104.2 and headscale version 0.29.5 in a three‑day lab setup. The researchers emphasised that the work was lab‑based and that Tailscale had not yet issued a security bulletin, CVE identifier or fixed version at publication.

Windisch’s tests indicate several privilege‑escalation paths: standard SSH port forwarding could cause the root‑level Tailscale process to access a root‑only system socket, enabling root access and rapid reading of the system password file; a second method involved replacing a root‑owned socket to allow direct root command execution; and a third bypassed a safety check to extend file access by racing path checks, ultimately enabling the theft of a device enrolment key to bring an attacker’s machine onto the network as a trusted server.

The report notes that these outcomes were observed only when Tailnet Lock was not enabled; networks with Tailnet Lock could mitigate some exposure, but the researchers assert that every SSH grant without proper controls could equate to root access on the target. Tailscale has acknowledged the disclosures and is reviewing affected setups while advising caution rather than a blanket disablement of Tailscale SSH.