ad-Orange
commented
3 months ago Hello, I agree with all changes for the exact reasons you mention except for Edit 2. For edit 2, instead of “is done by the WSCD”, I would write “must include the WSCD”. The justification of that is schemes like BBS# that combine WSCD “classic” signature (like ECDSA, ECSDSA) with an additional processing on the wallet instance side. Of course the private key used in the WSCD never leaves it, even in the BBS# kind of case. One could argue that if the processing on the wallet instance side is done by e.g. the keystore, then we’re looking at some kind of “multi-WSCD” processing but I think the depiction of it as a joint processing between the WSCD and the wallet instance is already good enough.
sander
The ARF describes when users must authenticate using “user authentication mechanisms” but says very little about what these mechanisms are. Many EUDI wallets for the Android platform will use a “Local WSCD” (described in Section 4.3) and when that WSCD architecture is chosen, by necessity the user authentication mechanism will be the Android device unlock (either a class-3 biometric, or the “lock screen knowledge factor” (LSKF))
We would like to suggest 3 ARF edits that will make this more clear.
Edit 1: Section 4.3
Original text:
Proposed new text:
Purpose of change: Makes it more clear that when a WSCD is used, the operating system itself is responsible for the user authentication mechanism.
Edit 2: Section 6.5.3
Original text:
Proposed new text:
Purpose of change: Makes it clear that user authentication is always the responsibility of the WSCD. User authentication performed outside of the WSCD cannot be trusted as it is done outside of the secure area
Edit 3: Section 4.2
Original text:
Proposed new text:
Purpose of change: The separation between WSCD and WSCA doesn’t map very well to the reality of an Android keystore. The Android Strongbox consists of both the secure element hardware and an applet running on it; it is therefore both a WSCD and a WSCA in one. However, the same applet is used for all applications on the device, so the sentence “One WSCA is associated with at most one Wallet Instance, and manages assets, such as keys, for this specific Wallet Instance.” does not make sense. We believe that getting rid of the WSCA concept and simply referring to the secure area as a WSCD (that can have many different architecture choices) will make the ARF easier to understand and to map to devices in the market today.