CVE-2018-1069

Related Vulnerabilities: CVE-2018-1069  

GlusterFS and NFS network filesystems rely on File System User ID and Group ID information in order to restrict access to file shares. However, it's possible to overwrite the Openshift restrictions on container UserId and GroupdId as they are not validated before being sent over the Openshift Network. An attacker could use the flaw to read and write any data on the network filesystem.

GlusterFS and NFS network filesystems rely on File System User ID and Group ID information in order to restrict access to file shares. However, it's possible to overwrite the Openshift restrictions on container UserId and GroupdId as they are not validated before being sent over the Openshift Network. An attacker could use the flaw to read and write any data on the network filesystem.

Find out more about CVE-2018-1069 from the MITRE CVE dictionary dictionary and NIST NVD.

CVSS v3 metrics

NOTE: The following CVSS v3 metrics and score provided are preliminary and subject to review.

CVSS3 Base Score 7.1
CVSS3 Base Metrics CVSS:3.0/AV:A/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H
Attack Vector Adjacent Network
Attack Complexity High
Privileges Required Low
User Interaction None
Scope Unchanged
Confidentiality High
Integrity Impact High
Availability Impact High

Affected Packages State

Platform Package State
Red Hat OpenShift Enterprise 3 Networking Will not fix

Acknowledgements

Red Hat would like to thank Andreas Skoglund (Basefarm AS) for reporting this issue.

Mitigation

If exposing shares with NFS or GlusterFS to Openshift Nodes use EgressNetworkPolicy, [1] to redirect outbound storage network traffic via an egress router. Alternatively if you're on AWS, EgressNetworkPolicy is not supported, you can use a Static IP, [2], for projects to force projects which need storage to connect from a NIC with a unique IP Address.

Once traffic is coming from a known IP address, not the Node IP, you can protect the NFS, or GlusterFS storage node with a firewall, or exports which only allows access from the Egress router, or from the Static IP.

An egress policy such as the one in the redirect-mode, [3], is an example of egress policy which would mitigate this vulnerability. An administrator would need to deploy one of these to the project using storage, as they need to run in privileged mode. This forces all traffic to the storage device through the egress router, which is a distinct IP. You can then configure the storage device to only access traffic from that source IP.

Be aware that during a docker build, the build container is not subject to EgressNetworkPolicy rules in versions prior to 3.7. Also the Static IP feature is only available as a Tech Preview from version 3.7.

An alternative mitigation if you're using Red Hat Gluster Storage is outlined here: https://github.com/gluster/gluster-kubernetes/blob/master/docs/design/tls-security.md

A semi-automatic namespace wide egress-IP is due to be released in OCP 3.11 which will make management of the EgressNetworkPolicy easier. It will only work for the multitenant and networkpolicy plugins.

[1] https://docs.openshift.com/container-platform/3.7/admin_guide/managing_networking.html#admin-guide-controlling-egress-traffic

[2] https://docs.openshift.com/container-platform/3.7/admin_guide/managing_networking.html#enabling-static-ips-for-external-project-traffic

[3] https://docs.openshift.org/latest/admin_guide/managing_networking.html#admin-guide-deploying-an-egress-router-pod