Tailworm Audits Tailscale Client v1.104.2 and headscale v0.29.5

Tailworm has published a security assessment of the Tailscale client agent v1.104.2 and headscale v0.29.5, the open source coordination server that implements the Tailscale control plane. The control plane is the coordinating layer that tells devices in a network how to find and authenticate one another. The report says its testing ran from 22 to 24 October 2026, and the document is dated 25 October 2026 (GMT).

The note is signed by Erica Windisch of Cyberdione Labs Corporation. Tailworm states that its researchers are not associated with, employed by, or contracted or paid by Tailscale Inc. It also states that all research was performed against open source software on the researchers' own systems and devices.

According to Tailworm, every finding was verified by reproduction in a lab, using unmodified, released software. The report labels its findings with internal audit identifiers in three series: T1 through T3, C1 through C4, and F-1 through F-5. These labels are the report's own scheme, and the report does not map them to a single severity ladder.

One finding is named specifically. Tailworm reports that tailscaled's embedded Tailscale SSH server runs as root, the highest-privilege account on a Unix-like system. For operators who run that embedded server, the process privilege level is the most directly actionable detail, because it determines what an attacker who reaches the service would inherit.

The report arrives against a separate vendor disclosure. On 13 October 2026, Tailscale published security bulletin TS-2026-012, which describes a bug in libtailscale's use of the SCM_RIGHTS API. SCM_RIGHTS is the Unix mechanism for passing open file descriptors, the handles a process uses to reach open files and sockets, between processes over a domain socket, a local communication channel. Tailscale advised users to update to Tailscale v1.104.2 or later. The client build Tailworm audited, v1.104.2, is the release Tailscale named in its advisory. The two documents are dated 13 October and 25 October respectively.

On disclosure timing, Tailworm states that it is the researcher's opinion that infrastructure is vulnerable today.

In my view, the practical reading for operators is the narrow one. The report names specific versions, a set of finding labels, and a privilege detail that can be checked against a running deployment. An operator running tailscaled with the embedded SSH server, or self-hosting headscale v0.29.5, has concrete configurations to review. Whatever weight one gives to the broader claim about infrastructure, the version and privilege specifics can be verified.

The broader context here is the method. Testing against released, unmodified builds, on the researchers' own hardware, lets maintainers, operators and other researchers repeat the work. That is a sound basis for the findings to be checked, challenged or confirmed, and it is the reason the reproduction claim matters more than any single label in the report. Open source networking software tends to improve when its released behavior can be examined this directly, and a published reproduction path gives maintainers a clear starting point for their own response.

Readers who run Tailscale or headscale in production will want to read the full report at tailworm.com, check the version they run against the bulletin's guidance, and review the privilege model of any embedded SSH deployment.