/

CVE-2026-72738

Severity:

CRITICAL RISK

(9.9)

OS Command Injection (CWE-78) in Dokploy Backup Endpoint

CVE-2026-72738

Received

Vulnerability type

Category:

OS Command Injection

CWE:

CWE-78

Impact:

Remote Code Execution

Severity:

CRITICAL RISK

(9.9)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

A critical OS command injection vulnerability in Dokploy's backup file-listing endpoint allows an authenticated user with backup:read permission to execute arbitrary commands on the Dokploy host. The issue becomes risky when user-controlled search input is incorporated into an rclone command without safe argument handling.

Affected context

The vulnerability occurs in Dokploy deployments using the backup.listBackupFiles tRPC endpoint. It requires a reachable backup-management interface, an authenticated account with backup:read permission, and processing of the search parameter by the backup router. Deployments that expose this functionality to untrusted or broadly authorized users may have greater exposure.

Exploit preconditions

  • Attack surface: Dokploy's backup.listBackupFiles tRPC endpoint
  • Required access: An authenticated account with backup:read permission
  • Trigger point: Processing the user-supplied search parameter during backup file listing
  • Why exploitation is possible: The parameter is incorporated into an rclone command executed through a shell-backed child-process API without sufficiently separating data from command syntax

Why this happens

The underlying bug is command construction from user-controlled input. Although the search value passes through path normalization, normalization alone does not establish that the value is safe for shell command execution. Interpolating that value into a command executed by child_process.exec() can allow shell metacharacters or command syntax to alter the intended operation, matching CWE-78: Improper Neutralization of Special Elements used in an OS Command.

Potential impact

  • An attacker may execute arbitrary commands on the Dokploy host.
  • The Dokploy service account or host environment could be compromised.
  • Sensitive application, backup, or cloud-storage credentials could be exposed.
  • Attackers could modify, delete, or exfiltrate hosted data.
  • The vulnerability could enable lateral movement or broader infrastructure compromise.

General mitigation guidance

Mitigation generally focuses on removing shell interpretation from backup operations and reducing the trust placed in request parameters.

Common defensive patterns include:

  • Using process-spawn APIs that pass executable arguments as an array without invoking a shell.
  • Applying strict allowlists and schema validation to search and path values.
  • Treating path normalization as a separate concern from command safety.
  • Using a dedicated rclone or storage-library interface rather than dynamically assembled shell commands.
  • Restricting backup-read permissions according to least privilege and separating them from administrative capabilities.
  • Running the service with minimal filesystem, network, and operating-system privileges.
  • Protecting backup endpoints with authorization boundaries, auditing, and rate controls.

How teams detect issues like this

Detection approaches can include:

  • Static analysis rules that identify shell-backed process execution with interpolated request values.
  • Code review checks for string concatenation or template interpolation in command arguments.
  • Tests that submit shell-sensitive input to endpoints handling paths, filenames, or search terms and verify that it remains data.
  • Runtime monitoring for unexpected child processes, command interpreters, or unusual rclone activity from the Dokploy service.
  • Policy enforcement that prohibits shell execution for externally influenced values unless a documented, reviewed exception exists.
  • Neutral automated code-review or software-composition tools that flag command-injection patterns.

Sources

SourceLink
github.comgithub.com/Dokploy/dokploy/commit/eeb6e7b8ea88e4b4b1fac8460755464100516ac9
github.comgithub.com/Dokploy/dokploy/releases/tag/v0.29.13
github.comgithub.com/Dokploy/dokploy/security/advisories/GHSA-5vjv-73wr-rf79

Frequently asked

questions

It may be exploitable through the Dokploy backup API when that interface is reachable and the attacker has the required authenticated permission. It is not generally an unauthenticated attack based on the described conditions.

The attacker needs an authenticated account authorized with backup:read permission. The practical exposure also depends on network access to the Dokploy instance and the privileges available to the affected service.

Avoid passing user-controlled values through shell interpretation. Prefer direct process APIs with structured arguments or a storage-library interface, combined with strict validation and allowlisted path semantics.

Treat all request parameters as untrusted, prohibit dynamic shell command construction in application code, enforce least-privilege authorization, and add static, dynamic, and runtime checks for unsafe process execution.

Find out what's already
exploitable in your codebase.