Package Instance

Privilege Escalation in Docker
Docker 1.0.0 uses world-readable and world-writable permissions on the management socket, which allows local users to gain privileges via unspecified vectors.

Path Traversal in Moby builder
util/binfmt_misc/check.go in Builder in Docker Engine before 19.03.9 calls os.OpenFile with a potentially unsafe qemu-check temporary pathname, constructed with an empty first argument in an ioutil.TempDir call.

NULL Pointer Dereference on moby image history
moby v25.0.0 - v26.0.2 is vulnerable to NULL Pointer Dereference via daemon/images/image_history.go.

Moby firewalld reload makes published container ports accessible from remote hosts
Moby is an open source container framework developed by Docker Inc. that is distributed as Docker Engine, Mirantis Container Runtime, and various other downstream projects/products. The Moby daemon component (dockerd), which is developed as [moby/moby](https://github.com/moby/moby) is commonly referred to as Docker, or Docker Engine.

Firewalld is a daemon used by some Linux distributions to provide a dynamically managed firewall. When Firewalld is running, Docker uses its iptables backend to create rules, including rules to isolate containers in one bridge network from containers in other bridge networks.

### Impact

The iptables rules created by Docker are removed when firewalld is reloaded using, for example "firewall-cmd --reload", "killall -HUP firewalld", or "systemctl reload firewalld".

When that happens, Docker must re-create the rules. However, in affected versions of Docker, the iptables rules that prevent packets arriving on a host interface from reaching container addresses are not re-created.

Once these rules have been removed, a remote host configured with a route to a Docker bridge network can access published ports, even when those ports were only published to a loopback address. Unpublished ports remain inaccessible.

For example, following a firewalld reload on a Docker host with address `192.168.0.10` and a bridge network with subnet `172.17.0.0/16`, running the following command on another host in the local network will give it access to published ports on container addresses in that network: `ip route add 172.17.0.0/16 via 192.168.0.10`.

Containers running in networks created with `--internal` or equivalent have no access to other networks. Containers that are only connected to these networks remain isolated after a firewalld reload.

Where Docker Engine is not running in the host's network namespace, it is unaffected. Including, for example, Rootless Mode, and Docker Desktop.

### Patches

Moby releases older than 28.2.0 are not affected. A fix is available in moby release 28.3.3.

### Workarounds
After reloading firewalld, either:
- Restart the docker daemon,
- Re-create bridge networks, or
- Use rootless mode.

### References
https://firewalld.org/
https://firewalld.org/documentation/howto/reload-firewalld.html

Privilege Elevation in runc
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.