A security researcher has published findings on weaknesses in Tailscale SSH, The Hacker News (opens in new tab) reported. According to the report, an attacker who already controls one device on an organization’s private Tailscale network could use the flaws to gain root access on other machines. Erica Windisch of Cyberdione Labs released the report, which she calls “Tailworm,” on Oct. 25. She said she tested Tailscale 1.104.2 and the open-source control server headscale 0.29.5 in a lab. No security bulletin, CVE identifier or fix had been issued when the report came out.
The attack has two prerequisites: SSH forwarding must be allowed, and the attacker must already control a device that is permitted to connect. Windisch lists 11 findings. “The mesh converts one ordinary foothold into fleet-wide privileged access at machine speed,” she wrote. She called the result “wormable,” a label The Hacker News (opens in new tab) said some researchers question.
In her lab tests, access rules written for specific device pairs and hardware-bound keys stopped the spread, and Tailscale’s Tailnet Lock feature blocked the final step. Headscale has no way to disable forwarding. Windisch’s proof-of-concept code has not been released. The Hacker News (opens in new tab) noted that an earlier related fix shipped in Tailscale 1.98.9.
Mike Shaver, Tailscale’s vice president of engineering, wrote on Reddit that the company had received the report and found issues it is fixing. He said Tailscale is “not recommending that everyone disable Tailscale SSH.” ThreatWire (opens in new tab) reported that Tailscale said the scenario “isn’t the same as an unauthenticated attacker being able to connect to arbitrary machines,” and that the company plans to publish an advisory.
According to ThreatWire (opens in new tab), one related code change (pull request #21931) has been merged into Tailscale’s repository and another (#21937) remains open. ThreatWire (opens in new tab) also said headscale 0.29.6 does not address the findings. No exploitation in the wild has been reported.
The report recommends several mitigations: access rules limited to specific device pairs, turning off SSH forwarding, removing app connector permissions, enforcing node-key expiry, and using Tailscale SSH’s check mode.