You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
physical: allow a user to touch a USB security key to select it
virtual: allow a user to view the connected USB security keys and click a button in the menu to start using it.
automatic: if a single USB security key is plugged in, then automatically start the ceremony on it.
Currently, we only support the first (physical). I think the latter two can be achieved by syncing a list of available devices to the backend, and then add a function on the frontend to select one of these devices by ID. That covers the virtual one. To achieve the automatic selection, the backend can choose, for example:
if no devices are connected at the start of the ceremony, and exactly one arrives, automatically select it
if exactly one device is present at the beginning of the ceremony, wait 1 second, then automatically select it.
This allows credentialsd not to mandate a specific pattern for selecting devices, and desktop environments can decide what's best for their users.
I think this means a few things:
We need a way to identify USB devices when they are plugged in. This can be as simple as an index
We need to query authenticator info for the name of the model for display purposes.
This might mean we'd need to cache the output of getInfo in the device struct.
We need to make devices a property of the ceremony that can be monitored for changes.
We have a pattern for receiving user input in the Enter PIN and credential selection flows. We can copy that for this input.
We may also want to consider updating our polling API to allow streaming in updates to the device list instead of polling, which would be more efficient and easier to track a stable list. (This is also a consideration for NFC, so I'll post this as a separate issue.)
I'd like to work on this. My plan is to start with the "virtual" mode:
In the USB flow, publish the connected devices to the UI with a simple per-ceremony index as the ID (for now, until libwebauthn can give a stable ID).
Add a device-selection input that follows the existing PIN and credential-selection pattern. Selecting a device starts the ceremony on it, and touching a key still works as it does today.
Then add the automatic mode in a follow-up PR: if exactly one device is present at the start or arrives first, select it after a short delay.
Model names would come later, once #257 and linux-credentials/libwebauthn#327 are settled. @msirringhaus, does this conflict with #231? I'm happy to base this on it or wait until it's merged. Does this split sound OK?
Activity
iinuwa commentedon Nov 9, 2026
For USB, maybe we could have libwebauthn expose a value based on the nonce returned by CTAPHID_INIT. Not sure if this maps to other transports.
kpanuragh commentedon Nov 11, 2026
I'd like to work on this. My plan is to start with the "virtual" mode:
Model names would come later, once #257 and linux-credentials/libwebauthn#327 are settled. @msirringhaus, does this conflict with #231? I'm happy to base this on it or wait until it's merged. Does this split sound OK?