CVE-2020-15110: Possible pod name collisions in jupyterhub-kubespawner
Impact What kind of vulnerability is it? Who is impacted?
JupyterHub deployments using:
- KubeSpawner <= 0.11.1 (e.g. zero-to-jupyterhub 0.9.0) and - enabled namedservers (not default), and - an Authenticator that allows: - usernames with hyphens or other characters that require escape (e.g. user-hyphen or user@email), and - usernames which may match other usernames up to but not including the escaped character (e.g. user in the above cases)
In this circumstance, certain usernames will be able to craft particular server names which will grant them access to the default server of other users who have matching usernames.
Patches Has the problem been patched? What versions should users upgrade to?
Patch will be released in kubespawner 0.12 and zero-to-jupyterhub 0.9.1
Workarounds Is there a way for users to fix or remediate the vulnerability without upgrading?
KubeSpawner
Specify configuration:
for KubeSpawner python from traitlets import default from kubespawner import KubeSpawner
class PatchedKubeSpawner(KubeSpawner): @default("podnametemplate") def defaultpodnametemplate(self): if self.name: return "jupyter-{username}-{servername}" else: return "jupyter-{username}"
@default("pvcnametemplate") def defaultpvcnametemplate(self): if self.name: return "claim-{username}-{servername}" else: return "claim-{username}"
c.JupyterHub.spawnerclass = PatchedKubeSpawner
Note for KubeSpawner: this configuration will behave differently before and after the upgrade, so will need to be removed when upgrading. Only apply this configuration while still using KubeSpawner ≤ 0.11.1 and remove it after upgrade to ensure consistent pod and pvc naming.
Changing the name template means pvcs for named servers will have different names. This will result in orphaned PVCs for named servers across Hub upgrade! This may appear as data loss for users, depending on configuration, but the orphaned PVCs will still be around and data can be migrated manually (or new PVCs created manually to reference existing PVs) before deleting the old PVCs and/or PVs.
References Are there any links users can visit to find out more?
For more information If you have any questions or comments about this advisory:
Open an issue in kubespawner Email us at security@ipython.org
Credit: Jining Huang
Other sources
In jupyterhub-kubespawner before 0.12, certain usernames will be able to craft particular server names which will grant them access to the default server of other users who have matching usernames. This has been fixed in 0.12.
jupyterhub-kubespawner could allow a remote attacker to bypass security restrictions, caused by improper validation of server names. By sending specially-crafted server names, an attacker could exploit this vulnerability to gain access to the default server of other users who have matching usernames.
— IBM
Affected Software
Remediation
Event History
Frequently Asked Questions
What is the vulnerability ID?
The vulnerability ID is CVE-2020-15110.
What is the severity of CVE-2020-15110?
The severity of CVE-2020-15110 is high with a value of 8.1.
Which software versions are affected by CVE-2020-15110?
Versions up to, but not including, 0.12 of Jupyterhub Kubespawner are affected by CVE-2020-15110.
How can this vulnerability be fixed?
The vulnerability has been fixed in version 0.12 of Jupyterhub Kubespawner.
Where can I find more information about CVE-2020-15110?
More information about CVE-2020-15110 can be found in the references: [GitHub Commit](https://github.com/jupyterhub/kubespawner/commit/3dfe870a7f5e98e2e398b01996ca6b8eff4bb1d0) and [GitHub Security Advisories](https://github.com/jupyterhub/kubespawner/security/advisories/GHSA-v7m9-9497-p9gr).