docker

runc through 1.0-rc6, as used in Docker before 18.09.2 and other products, allows attackers to overwrite the host runc binary (and consequently obtain host root access) by leveraging the ability to execute a command as root within one of these types of containers: (1) a new container with an attacker-controlled image, or (2) an existing container, to which the attacker previously had write access, that can be attached with docker exec. This occurs because of file-descriptor mishandling, related to /proc/self/exe.

HandleRequestAsync in Docker for Windows before 18.06.0-ce-rc3-win68 (edge) and before 18.06.0-ce-win72 (stable) deserialized requests over the \\.\pipe\dockerBackend named pipe without verifying the validity of the deserialized .NET objects. This would allow a malicious user in the "docker-users" group (who may not otherwise have administrator access) to escalate to administrator privileges.

Docker before 1.3 does not properly validate image IDs, which allows remote attackers to redirect to another image through the loading of untrusted images via 'docker load'.

Docker before 1.5 allows local users to have unspecified impact via vectors involving unsafe /tmp usage.

Docker Engine 1.12.2 enabled ambient capabilities with misconfigured capability policies. This allowed malicious images to bypass user permissions to access files within the container filesystem or mounted volumes.

libcontainer/user/user.go in runC before 0.1.0, as used in Docker before 1.11.2, improperly treats a numeric UID as a potential username, which allows local users to gain privileges via a numeric username in the password file in a container.