Skip to main content
Introducing Device Trust
Introducing Device Trust

Length: 01:35

Enforce Device Trust

Supported Resources

Device trust only supports SSH, Database and Kubernetes resources. Support for other resources is planned for upcoming Teleport versions.

Resources protected by the device mode "required" will enforce the use of a trusted device, in addition to establishing the user's identity and enforcing the necessary roles. Furthermore, users using a trusted device leave audit trails that include the device's information.

Device Trust enforcement can be configured with the following three modes of operation, represented by the device_trust_mode authentication setting:

  • off - disables device trust. Device authentication is not performed and device-aware audit logs are absent.
  • optional - enables device authentication and device-aware audit, but does not require a trusted device to access resources.
  • required - enables device authentication and device-aware audit. Additionally, it requires a trusted device for all SSH, Database and Kubernetes connections.

Prerequisites

  • A running Teleport Enterprise cluster. For details on how to set this up, see the Enterprise Getting Started guide.

  • The Enterprise tctl admin tool and tsh client tool version >= 14.3.33. You can download these tools by visiting your Teleport account. You can verify the tools you have installed by running the following commands:

    $ tctl version
    # Teleport Enterprise v14.3.33 go1.21

    $ tsh version
    # Teleport v14.3.33 go1.21
  • To enroll a macOS device, you need:
  • To enroll a Windows device, you need:
    • A device that includes a TPM with 2.0 support.
    • tsh installed on the device. Download the Windows tsh installer.
    • Credentials for a user with administrator privileges for the device. This is only required during enrollment.
    • A Teleport version newer than v13.1.2.

Role-based trusted device enforcement

Role-based configuration enforces trusted device access at the role level. It can be configured with the spec.options.device_trust_mode option and applies to the resources in its allow rules. It works similarly to require_session_mfa.

v13.3.6+

Teleport version 13.3.6 and above has the preset require-trusted-device role. Make sure you update the "allow" rules in the role according to your requirements.

To enforce authenticated device checks for a specific role, update the role with the following:

kind: role
version: v7
metadata:
name: require-trusted-device
spec:
options:
# require authenticated device check for this role
+ device_trust_mode: "required" # add this line
allow:
logins: ['admin']
kubernetes_groups: ['edit']
node_labels:
'*': '*'
...

$ tctl create -f device-enforcement.yaml

Cluster-wide trusted device enforcement

Cluster-wide configuration enforces trusted device access at the cluster level. Enterprise clusters run in optional mode by default. Changing the mode to required will enforce a trusted device for all SSH, Database and Kubernetes accesses.

Web UI

The Web UI is not capable of trusted device access. Only tsh and Teleport Connect are able to fulfill device mode required.

To enable device mode required update your configuration as follows:

Create a cap.yaml file or get the existing configuration using tctl get cluster_auth_preference:

kind: cluster_auth_preference
version: v2
metadata:
name: cluster-auth-preference
spec:
type: local
second_factor: "on"
webauthn:
rp_id: teleport.example.com
device_trust:
+ mode: "required" # add this line

Update the configuration:

$ tctl create -f cap.yaml
cluster auth preference has been updated

You can also edit this configuration directly:

$ tctl edit cluster_auth_preference

Once the config is updated, SSH, Database and Kubernetes access without a trusted device will be forbidden. For example, SSH access without a trusted device fails with the following error:

$ tsh ssh ip-172-31-35-170
ERROR: ssh: rejected: administratively prohibited (unauthorized device)
Trusted Clusters

It is possible to use trusted clusters to limit the impact of device mode required. A leaf cluster in mode required will enforce access to all of its resources, without imposing the same restrictions to the root cluster. Likewise, a root cluster will not enforce device trust on resources in leaf clusters.

Locking a device

Similar to session and identity locking, a device can be locked using tctl lock.

Locking blocks certificate issuance and ongoing or future accesses originating from a locked device. Locking a device only works if device trust is enabled and if the device is enrolled to Teleport.

Find a device ID to lock:

$ tctl devices ls
Asset Tag OS Enroll Status Device ID
------------ ----- ------------- ------------------------------------
C00AA0AAAA0A macOS enrolled 9cdfc0ad-64b7-4d9c-this-is-an-example

Lock a device:

$ tctl lock --device=9cdfc0ad-64b7-4d9c-this-is-an-example --ttl=12h
Created a lock with name "5444970a-39a0-4814-968d-e58b4a8fa686".

Now, if a user on that device tries to access an SSH server for example, Teleport will deny access:

$ tsh ssh ip-172-31-35-170
ERROR: ssh: rejected: administratively prohibited (lock targeting Device:"9cdfc0ad-64b7-4d9c-this-is-an-example" is in force)

Troubleshooting

"binary missing signature or entitlements" on tsh device enroll

A signed and notarized tsh binary is necessary to enroll and use a a trusted device. Download the macOS tsh installer to fix the problem.

"unauthorized device" errors using a trusted device

A trusted device needs to be registered and enrolled before it is recognized by Teleport as such. Follow the registration and enrollment steps and make sure to tsh logout and tsh login after enrollment is done.

Auto enrollment not working

Auto-enrollment ceremonies, due to their automated nature, are stricter than regular enrollment. Additional auto-enrollment checks include:

  1. Verifying device profile data, such as data originated from Jamf, against the actual device
  2. Verifying that the device is not enrolled by another user (auto-enroll cannot take devices that are already enrolled)

Check you audit log for clues: look for failed "Device Enroll Token Created" events and see the "message" field in the details (auto-enroll audit log details available since Teleport v14.3.33).

If you suspect (1) is the issue, compare the actual device against its inventory definition (tsh device collect executed in the actual device vs tctl get device/<asset_tag>). Tweaking the device profile, manual enrollment or waiting for the next MDM sync may solve the issue.

If you suspect (2), you can unenroll the device using tctl edit device/<asset_tag> and changing the "enroll_status" field to "not_enrolled".

Next steps