Tailscale Warns of 11 ‘Tailworm’ Flaws Including Root SSH Escalation

Tailscale’s security team is working to release a patched software update by the end of Tuesday, Oct. 27, to address 11 security vulnerabilities—collectively dubbed “Tailworm”—that could allow authorized attackers on a targeted tailnet to gain root code execution, steal cloud credentials, or crash control servers, according to BleepingComputer.

Discovered and verified in a lab environment by Cyberdione Labs founder Erica Windisch on unmodified releases of Tailscale v1.104.2 and the self-hosted headscale v0.29.5 control server, the flaws span five components:

1. Tailscale SSH (tailscaled): Unix-socket forwarding (ssh -R/-L) performs path checks and file operations as root without dropping privileges, allowing an approved device with an SSH policy rule to execute code as root, capture enrollment keys, or delete root files.
2. Tailscale App Connectors (conn25, experimental): Translates traffic to attacker-chosen destinations without consulting network ACLs, enabling access to cloud metadata, internal LANs, and Kubernetes cluster services.
3. Self-Hosted headscale Control Server: Unauthenticated memory growth and cache eviction can trigger a control-server crash loop and block new device onboarding.
4. Tailscale DERP Client: Unbounded reads from a hostile relay can exhaust client-agent memory.
5. tailscale serve: TCP forwarding to a Unix socket lacks an HTTP guard and contains a per-mount authorization race.

Mitigation Guidance From Tailscale Engineering

Tailscale VP of Engineering Mike Shaver said the company does not recommend that everyone disable Tailscale SSH entirely. Instead, until Tuesday’s patched release is applied, administrators should tighten SSH access policies and disable SSH forwarding if unneeded by setting TS_SSH_DISABLE_FORWARDING=1 in the tailscaled service environment and restarting the service.