Re: [bitcoin-dev] URI scheme for signing and verifying messages

[Luke Dashjr <luke@dashjr.org>]

On Tuesday, September 15, 2015 10:49:36 AM Arthur - bitcoin-fr.io wrote:
> September 15 2015 6:04 AM, "Luke Dashjr" <luke@dashjr.org> wrote:
> > I think probably the whole signed message thing needs to be rethought.
> > The most common "uses" today seem to be insecure cases that it doesn't
> > actually work in: people trying to prove ownership of bitcoins and/or
> > that they sent bitcoins (current signed messages can do neither).
> > Ideally, whatever the new method is should also avoid using the same key
> > as for signing transactions, since the public key is technically private
> > information. Furthermore, since addresses are semi-deprecated (by the
> > payment protocol), I'm not sure it makes sense to do this without
> > designing an entire authentication system, which may be rather complex.
> > 
> > Luke
> 
> My proposal is about the current signing process (which exists event it
> it's not perfect) but it could also work with a new signing message system
> tomorrow. It more about give users an easier way to access existing tools
> than the "sign message thing" itself.

One of my concerns is that making the existing signatures even easier will 
cause incompatible uses to become more prolific and accepted, increasing the 
overall risk. Hence my recommendation to satisfy these clearly-existing use 
cases with a safe signature *first*.

> BTW I'm aware of privacy issues, but could you elaborate on why the use
> case your are referring to doesn't actually work?

The signed message proves that the person who *receives* payment with the 
address agrees to a given message/contract.

It is therefore appropriate and a best practice for web wallet providers 
(inherent problems with webwallets aside) to allow users to sign messages 
with their deposit addresses. When bitcoins are received by this address, the 
transaction creates a low-level UTXO representing the bitcoins *in the 
wallet*, but this UTXO is not associated with the address itself. Therefore, 
it is entirely possible that this UTXO remains unspent/valid on the 
blockchain even after the user in question has spent their entire balance at 
the webwallet and therefore such a signature proves only that they once 
received the payment, but *not* that they presently still have the bitcoins 
received.

Furthermore, it is proper for the UTXO to be redeemed at a low-level by the 
wallet when an entirely unrelated user is sending a transaction. In such a 
circumstance, the original recipient of the bitcoins would still be able to 
sign a message, even though they have nothing to do with nor any right to the 
goods/services purchased with the transaction redeeming that UTXO.

> Here are a use of
> bitcoin signatures ( https://bitcointalk.org/index.php?topic=497545.0 ) to
> speak about a real case.

Yes, there are a few good use cases for the current signed messages, but they 
appear to be a minority at the moment. I suspect implementing any URI-based 
signing would actually make them more difficult as well, since it is 
additional code on the requester's part.

Luke