Multi-Device Pairing Considerations

Shared MUAA Messaging Archive

characteristics/requirements of of what SMAs need to provide:

  • a SMA can be implemented on top of IMAP commands
  • is used to synchronize states between MUAAs. We use “MUAAs” to indicate a particular MUA/Account combination because synchronization happens betweens accounts managed by different MUAs.
  • is used to send and receive messages between MUAAs (concurrently), for example pairing requests, initial Autocrypt setup (of first MUAA), updates to received remote Autocrypt encryption keys.
  • A MUAA needs to be able to detect if there is any other MUAA
  • messages are not (necessarily) human readable and don’t appear in the regular inbox.
  • probably: size of SMA should not grow linearly with number of incoming/outgoing mails, for example messages that have been processed by a MUA must be deleted
  • there should be a policy/expiry of messages for MUAAs which don’t exist/are not alive anymore
  • we only require from IMAP servers that they handle first level folders (subfolders are not necessary)
  • there is a header in the messages stored in these folders, indicating that the message is an SMA message.

implementation on top of IMAP, pairing happy path

Let’s suppose we have a first MUAA. It doesn’t find an _Autocrypt_SMA announcement folder so it will do the following:

  • create a random new number “1” which we call MUAA-ID.
  • create an _Autocrypt_SMA “announcements” folder and append some MUAA description message, most notably the MUAA-ID
  • create an inbox folder _Autocrypt_SMA_1 where other MUAAs will be able to send/drop messages.

If now another MUAA is added:

  • create a random new number “27” as MUAA-ID.
  • discover the _Autocrypt_SMA folder exists and read all of its messages, discover that there is an 1 MUAA
  • create an inbox folder _Autocrypt_SMA_27 where other MUAAs will be able to send/drop messages.
  • append a new MUAA description message to _Autocrypt_SMA
  • append a pairing request message to the “1” inbox (_Autocrypt_SMA_1).

The MUAA “1” will then:

  • discover “27” from the new message in the announcement folder _Autocrypt_SMA
  • read the pairing request message from its own _Autocrypt_SMA_1 inbox
  • process the pairing request and send a pairing accept message to “27” by appending it to the _Autocrypt_SMA_27 folder.
  • delete the pairing request message from its own _Autocrypt_SMA_1 folder.

Note

In this happy path example we are not prescribing the precise pairing procedure, merely give an example how bootstrapping into a multi-MUA setting works. It is unclear whether a centrally shared keyring as an IMAP folder is viable (synchronization between MUAs, “merge conflict” between state, deleting message might be a problem, encrypted “broadcast” to all my MUAAs)

Todo

Critically consider how the multiple Autocrypt folders show in user interfaces. It might be better to depend on sub folders.

Todo

Crically consider end-to-end encryption for MUAA messages.

Todo

Consider how to force remove devices through IMAP folder deletion or something.

types of inter-MUAA unicast messages

Difficult to reason about when we don’t know what we really want to do (cryptographic protocol wise)

ID announcement

pairing messages

  • Some authenticated key exchange so later messages between MUAAs can be encrypted
  • Shared private key so messages encrypted to the account’s public key can be encrypted and outgoing mail can be signed

remote key updates

  • notify other MUAAs that you add to or change an entry to your keyring

Note

Why not use IMAP METADATA instead of specially-named folders?

We ultimately want Autocrypt to be more generic than IMAP, to make it clear how other mail-checking protocols could work (e.g. MAPI, webmail interfaces) as long as they offer some sort of namespaced shared storage. Using IMAP METADATA would tie Autocrypt more tightly to IMAP, and would also limit the number of IMAP implementations that Autocrypt-enabled MUAs could connect to (METADATA is not widely supported by today’s IMAP server implementations).

If we wanted Autocrypt to use METADATA where it was available on the server, but allow for fallback to normal folders for IMAP servers that don’t support METADATA, then we’d be adding an implementation requirement for MUAs that might not already know how to use the METADATA extension, which makes adoption harder.

And without initially requiring it for MUAs, we don’t see a way to transition once non-METADATA capable MUAs exist in the wild, either, since lockout and sync become difficult to do. So we don’t see a good story for METADATA deployment, sadly, despite it targeting our use case fairly neatly.

See also earlier discussion about IMAP METADATA.