Re: [bitcoin-dev] BIP proposal: Timelocked address fidelity bond for BIP39 seeds

[Chris Belcher <belcher@riseup.net>]

Hello ZmnSCPxj,

You say "A taker can be a surveillor as well", as though that's simple 
and easy to achieve. In reality there are many defenses against that.

Defending against the attack of a malicious taker aborting at the last 
step is the purpose of the podle commitments which joinmarket has 
implemented since 2016. This was in response to this attack actually 
taking place. Another important point is that this attack cant happen 
secretly, it is very obvious to everyone operating a maker that it 
happens. The podle defense means that an attacker doing this will 
constantly have to spend money on miner fees to create new UTXOs. Here's 
a writeup with links to other blog posts about the whole thing: 
https://gist.github.com/chris-belcher/00255ecfe1bc4984fcf7c65e25aa8b4b

As well as podle as mitigation, the multiple mixdepths in the joinmarket 
wallet also helps a lot because it's not trivial for an attacker to 
actually learn all the UTXOs in all 5 mixdepths, which is necessary for 
the attack to work.

Mitigation in Teleport works in a slightly different way: takers can 
only see UTXOs or transactions belonging to the maker once they have 
already gotten their own transaction confirmed. So if they were to abort 
the protocol early they would not only have spent miner fees but also 
waste their own time waiting for the OP_CSV timeout.

It's worth remembering that the fidelity bond UTXOs are not linked to 
any resulting coinjoin or coinswaps on-chain.

Yes linking the two identities (joinmarket maker and teleport maker) 
together slightly degrades privacy, but that has to be balanced against 
the privacy loss of leaving both systems open to sybil attacks. Without 
fidelity bonds the two systems can be sybil attacked just by using about 
five-figures USD, and the attack can get these coins back at any time 
when they're finished.

Regards
CB

On 13/05/2022 13:44, ZmnSCPxj wrote:
> Good morning Chris,
> 
>> Hello waxwing,
>>
>>> A user sacrifices X amount of time-value-of-money (henceforth TVOM)
>>
>> by committing in Joinmarket with FB1. He then uses the same FB1 in
>> Teleport, let's say. If he gets benefit Y from using FB1 in Joinmarket,
>> and benefit Z in Teleport, then presumably he'll only do it if
>> (probabilistically) he thinks Y+Z > X.
>>
>>> But as an assessor of FB1 in Joinmarket, I don't know if it's also
>>
>> being used for Teleport, and more importantly, if it's being used
>> somewhere else I'm not even aware of. Now I'm not an economist I admit,
>> so I might not be intuit-ing this situation right, but it fees to me
>> like the right answer is "It's fine for a closed system, but not an open
>> one." (i.e. if the set of possible usages is not something that all
>> participants have fixed in advance, then there is an effective Sybilling
>> problem, like I'm, as an assessor, thinking that sacrificed value 100 is
>> there, whereas actually it's only 15, or whatever.)
>>
>>
>> I don't entirely agree with this. The value of the sacrifice doesn't
>> change if the fidelity bond owner starts using it for Teleport as well
>> as Joinmarket. The sacrifice is still 100. Even if the owner doesn't run
>> any maker at all the sacrifice would still be 100, because it only
>> depends on the bitcoin value and locktime. In your equation Y+Z > X,
>>
>> using a fidelity bond for more applications increases the
>> left-hand-side, while the right-hand-side X remains the same. As
>> protection from a sybil attack is calculated using only X, it makes no
>> difference what Y and Z are, the takers can still always calculate that
>> "to sybil attack the coinjoin I'm about to make, it costs A btc locked
>> up for B time".
> 
> I think another perspective here is that a maker with a single fidelity bond between both Teleport and Joinmarket has a single identity in both systems.
> 
> Recall that not only makers can be secretly surveillors, but takers can also be secretly surveillors.
> 
> Ideally, the maker should not tie its identity in one system to its identity in another system, as that degrades the privacy of the maker as well.
> 
> And the privacy of the maker is the basis of the privacy of its takers.
> It is the privacy of the coins the maker offers, that is being purchased by the takers.
> 
> 
> A taker can be a surveillor as well, and because the identity between JoinMarket and Teleport is tied via the single shared fidelity bond, a taker can perform partial-protocol attacks (i.e. aborting at the last step) to identify UTXOs of particular makers.
> And it can perform attacks on both systems to identify the ownership of maker coins in both systems.
> 
> Since the coins in one system are tied to that system, this increases the information available to the surveillor: it is now able to associate coins in JoinMarket with coins in Teleport, via the shared fidelity bond identity.
> It would be acceptable for both systems to share an identity if coins were shared between the JoinMarket and Teleport maker clients, but at that point they would arguably be a single system, not two separate systems, and that is what you should work towards.
> 
> 
> Regards,
> ZmnSCPxj