On this page (7 sections)
Dokploy, a self-hosted deployment platform, has a critical OS command injection flaw tracked as CVE-2026-129433. According to the GitHub advisory, versions before 0.31.1 let an authenticated organization member run shell commands as root when scheduled backups run, and from there take over the host through the mounted Docker socket.
If you run Dokploy for your own sites, for clients or for a team, this matters most when more than one person has an account. The advisory describes versions before 0.31.1 as affected, so plan the update now.
What happened
The advisory was published on November 14, 2026. It describes a flaw in a function named getComposeContainerCommand(). An authenticated organization member can supply a serviceName value containing shell metacharacters and store it through the backup.create or backup.update tRPC procedures. The value is later used when building a command, so the injected text runs when a scheduled backup executes.
The advisory states that the commands run as root and that the attacker can take over the host through the mounted Docker socket. Only one independent publisher covers this so far. The NVD entry repeats the same description, so it is not separate confirmation, and the sources reviewed do not report any real-world attacks.
Technical details in plain terms
OS command injection (CWE-78) happens when software builds a shell command from text a user controls without neutralizing special characters. Characters such as semicolons or backticks can then end the intended command and start another one. Here, the user-controlled text is the service name attached to a compose backup configuration.
Two details make this worse than a typical injection bug:
- The malicious value is stored first and executed later, so the command runs when the backup job fires, not when the request is made. Logs from the request alone may look harmless.
- The job runs as root and the Docker socket is mounted. Access to that socket is effectively control of the host, because it lets a process start containers with broad access.
The CVSS v4 metrics in the advisory show a network attack vector, low complexity, no attack requirements, low privileges required and no user interaction, with high impact on confidentiality, integrity and availability for both the vulnerable and subsequent systems. The project's code is public, including a referenced commit and pull request 5908.
Who is affected
Any Dokploy instance running a version older than 0.31.1 is in scope. The attacker needs a valid account that is a member of an organization, so the realistic exposure is:
- Teams or agencies where several people, contractors or clients have Dokploy logins.
- Instances with open or loosely controlled sign-up or invitations.
- Setups where one stolen or reused member password would otherwise be a limited problem. With this flaw, a low-privilege login can become root on the host.
- Servers that host other workloads next to Dokploy, since host takeover exposes everything on the machine.
A single-user instance with strong credentials has lower exposure, but updating is still the right call.
How to check your exposure
First find out which version you run. The version is normally visible in the Dokploy dashboard; confirm it is 0.31.1 or newer. Then review who can log in and what backups exist.
On the host, this read-only command lists running containers with their images and tags, which helps you spot the Dokploy image version in use (look for a tag below 0.31.1):
docker ps --format '{{.Names}}\t{{.Image}}\t{{.Status}}'Next, review the backup configurations in the dashboard for compose services. Look for service names that contain anything other than a normal service name, such as semicolons, backticks, dollar signs, pipes or ampersands. Treat any such value as suspicious.
What to do now
- Update Dokploy to 0.31.1 or later, following the v0.31.1 release notes and your usual upgrade method.
- Before or right after updating, list all organization members and remove accounts you do not recognize or no longer need.
- Inspect existing compose backup entries and edit or delete any whose serviceName contains shell metacharacters.
- Rotate passwords for all Dokploy users, and enable any extra sign-in protection your setup offers.
- If you find a suspicious serviceName or cannot rule out misuse, treat the host as possibly compromised: rotate secrets, API tokens, registry credentials and deploy keys that were stored on or reachable from it.
- Restrict network access to the Dokploy dashboard, for example to a VPN or an allowlist of admin IPs.
- Consider whether the Docker socket mount is needed for every workload on that host, and keep unrelated services on separate machines.
Detection and monitoring afterwards
Because execution happens at backup time, look at what ran around your backup schedule. On the host, this read-only command lists all containers, including stopped ones, with their creation time, which can reveal containers you did not start:
docker ps -a --format '{{.Names}}\t{{.Image}}\t{{.CreatedAt}}'Also check for unexpected new users, SSH keys, cron entries and outbound connections on the host. If your Dokploy panel sits behind a public hostname, keep an eye on its certificate and DNS records for unexpected changes; Webforta Monitor can alert on certificate and DNS changes, though it does not detect this flaw itself. For general hardening of exposed admin panels, see small business website security.
Timeline
| Date | Event |
|---|---|
| Nov 14, 2026 | CVE-2026-129433 published by NVD and added to the GitHub Advisory Database |
| Nov 14, 2026 | Advisory last updated in the GitHub Advisory Database |
The sources do not give a disclosure date, a reporter, or the release date of 0.31.1, so none are stated here.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
