Docker Hub flaw let attackers impersonate users and alter images

Docker disclosed a critical vulnerability in Docker Hub that could let attackers use a personal access token (PAT) to access another user’s container repositories. The flaw, tracked as CVE-2026-105445, could allow image pulls, pushes, or deletions, depending on the token’s scope and the target user’s permissions.

The issue occurred when an attacker created an account and PAT, converted the account into an organization, and named a target user as its owner. Docker Hub reassigned the PAT to that owner without changing its secret or seeking the owner’s consent. The token could then authenticate as the target at the registry level, potentially allowing images to be published under a legitimate name. Docker rated the issue critical and described it as a software supply chain risk.

Docker said the flaw did not provide access to Docker Hub’s website or let attackers change passwords, email addresses, or two-factor settings. It was introduced on August 26, 2024. Docker received a report on September 30, 2026, confirmed the issue, and deployed a production fix less than four hours later.

Docker said it found no evidence of exploitation after examining 304,000 pushes made with transferred tokens and reviewing high-risk public images. It notified affected users and organization owners on October 2, then deactivated transferred PATs on October 8, except those whose owners explicitly asked to keep them. Docker said users who received no notification should assume they were not affected and need take no action. The platform no longer transfers PATs during conversion and asks for consent before naming someone an organization owner.