CVE-2026-44883: Portainer: JWT accepted in URL query leaks tokens to logs and referers

Synopsis

This is a token-leakage vulnerability in Portainer Community Edition (versions 2.33.0 through several release lines before 2.33.8, 2.39.2, and 2.41.0). Portainer's authentication middleware accepts JWT bearer tokens passed as a plain URL query parameter (?token=), and because URLs are recorded in reverse-proxy access logs, browser history, and HTTP Referer headers, those tokens are exposed to anyone who can read those logs or to external sites the user navigates to afterward. A leaked token carries the full privileges of the issuing user for up to eight hours, enabling account takeover, container manipulation, and service disruption. A patched-image rebuild is available on HarborGuard for environments running an affected version.

HarborGuard Coverage

Exploit Conditions

• Network reachabilityRequired

  The attacker must reach the Portainer API service over the network to harvest a token from logs or to receive it via an HTTP Referer header from an external site the victim navigates to.

• AuthenticationNot required

  No account or credentials are needed by the attacker; the JWT is harvested passively from log access or an outbound Referer header rather than through an authenticated channel.

• Victim interactionRequired

  A legitimate Portainer user must perform an action that triggers the ?token= query parameter (such as using the container attach, exec, or pod shell feature), generating a log entry or Referer header that leaks the token.

• Attack complexityDetail

  Exploitation depends on environmental factors: the attacker must have access to reverse-proxy logs, browser history, or be in a position to observe outbound Referer headers, making reliable exploitation conditional on those access paths being present.

Blast Radius

• A leaked token grants the full API privileges of the victim user for up to eight hours, allowing the attacker to read all container configurations, secrets, and environment variables visible to that user.
• The attacker can execute arbitrary commands inside running containers or attach to container shells, enabling direct tampering with running workloads.
• Any container or pod the victim had exec or attach rights on can be stopped, deleted, or reconfigured by the attacker using the stolen token.
• The attacker can crash or restart containers under the victim's scope, causing service disruption to any workloads managed through that Portainer instance.