Signal K Server leaks admin keys via log download, enabling full takeover
Anyone on the internet can download the server's log files without logging in. Those files contain a secret key that lets an attacker pretend to be an administrator and take full control of the server.
| Severity | CriticalCVSS 9.3 of 10 |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | YesSignal K Server 2.33.0 or later · since 28 Oct 2026 |
| How sure is this? | Official sourceBacked by an official source |
How the attack works simplified
- 1Remote attackerNo login or account needed
- 2Request logsCalls /skServer/ziplogs endpoint
- 3Extract secret keysecurity.json in ZIP holds JWT key
- 4Forge admin tokenSigns tokens with stolen secret
- 5Create admin accountPersistent access to server
- ResultAttacker gains full admin control
What happened
Signal K Server versions before 2.33.0 expose the /skServer/ziplogs endpoint without authentication; the downloaded ZIP includes security.json, which contains the JWT secretKey used to sign admin tokens.
Are you affected?
You are affected if you run Signal K Server earlier than 2.33.0 and its web interface or port is reachable from the internet or an untrusted network.
Signal K Server versions before 2.33.0
What to do
- Update Signal K Server to version 2.33.0 or later
- Restrict network access to the server so it is not exposed to the internet
- Review admin accounts and rotate the JWT secret key after patching if the server was exposed
- Check logs for unexpected requests to /skServer/ziplogs
Timeline
- 28 Oct 2026, 22:17 UTCPublicly disclosed
- 28 Oct 2026, 22:17 UTCReported by NVD · NVD
- 28 Oct 2026, 22:18 UTCPicked up by Zero Radar
Sources 1 report, oldest first
Official vendor, government CERT or advisory database
- CVE-2026-118690NVD · 28 Oct 2026, 22:17 UTC
Technical details
- ID
- CVE-2026-118690
- CVSS
- 9.3 / 10
- Disclosed
- 28 Oct 2026, 22:17 UTC
- On Zero Radar
- 28 Oct 2026, 22:18 UTC (2 min after disclosure)
- Last change
- 28 Oct 2026, 22:18 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works