Re: [bitcoin-dev] Proof of reserves - recording

[Billy Tetrud <billy.tetrud@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 10351 bytes --]

@ZmnSCPxj
>  a thief (or "thief") who has gotten a copy of the key can sign a
transaction that spends it, one second after the proof-of-reserves is made.

Sure, but anyone can easily see that transaction happen and can discount
that part of the attestation. The same isn't true on lightning.

> *knowledge* is easily copyable.

Knowledge isn't even really needed. Custodians could collude to sign each
other's balance attestations. However, if the network of proof of reserves
is cohesive, people could validate that addresses (and their balances)
aren't shared by anyone that's part of that PoR network.

> There is no way to prove that there is no alternate copy of the privkeys

Well there are only really 2 cases:

1. Only one entity has the keys
2. Two entities have the keys and trust each other

In case 1, the attestation is accurate. In case 2, its really another way
of saying case 1: you can view the two trusting entities as a single
entity. In the case there are 2 non-trusting entities, its very likely that
one would steal from the other, or that they would be very uncomfortable
with the situation such that it wouldn't last long.

But you can't prove that someone won't steal the reserves. You can't prove
cryptographically that the reserves aren't promised to someone else in a
legal contract (unless of course you make the proof of reserves system
legally binding). But this is why anyone that recommends PoR also
recommends independent audits to attest that the PoR actually matches
reality.

So yes, there are limitations, but I don't think that means that PoR isn't
worth doing. Is that what you're implying?

> we can show evidence that we live in a landlocked city far from any
lakes, seas, or rivers

I think that's the idea, if I understand you correctly.

@Eric
Auditability Fallacy
<https://github.com/libbitcoin/libbitcoin-system/wiki/Auditability-Fallacy>

> A solvency audit requires simultaneous (atomic) proof of both the full
amount of the asset held by a custodian and the securities issued against
it.

> in the case where the security is issued on a distinct public chain the
atomicity requirement is not satisfied.

I think what its saying is that you can't get atomicity of both the
security and the reserve. While this is true, what you can get is a system
where the order of events can be established to a degree of precision. Ie,
you can know that between reserve-snapshot A and B, the balances add up to
X. Each user can validate that their balance was indeed that value between
A and B. With reserve snapshots and balance snapshots frequent enough, this
can allow reasonably high accuracy of estimated solvency. However, it does
seem clear that perfect accuracy is not possible.

> Historically it has not been difficult to detect such deviations. The
difficulty arises in stopping them.

I disagree here that it has not been difficult to detect deviations
(insolvency). I mean, "difficulty" isn't the right word. These things
always become clear eventually. But I think its important to detect
insolvency quickly. Historically insolvency has certainly not been detected
quickly. Insolvency is instead practically perpetual, and the question is
only how insolvent and when will it explode?

I'm of the opinion that you can't prevent insolvency. Places will have
money troubles and will try to cover it up, since usually there is no
downside (admitting insolvency can lead to bankrupcy, and failure to
conceal insolvency has the same result - so why not try to conceal it and
hope you can shore it up). However, its important that people know the
institutions they have their money in are insolvent, or to what degree they
are. If that information were well tracked, it could become clear over time
that a 10% insolvent company rarely goes out of business, but a 20%
insolvent company usually does. Then people can have parameters where
they're ok with a certain measurable degree of insolvency, but react
swiftly and strongly when a company is too reckless. Currently the amount
of recklessness any given company engages in is basically a company secret
that their clients don't have insight into. PoR would greatly help this I
think. You don't think so?

On Mon, Jul 5, 2021 at 10:10 PM Eric Voskuil <eric@voskuil.org> wrote:

> https://github.com/libbitcoin/libbitcoin-system/wiki/Auditability-Fallacy
>
> On Jul 5, 2021, at 21:54, ZmnSCPxj <ZmnSCPxj@protonmail.com> wrote:
>
> Good morning Billy,
>
>
>
>   The two participants in the channel can sign a plaintext containing
> their node pubkeys and how much each owns
>
>
> Sure, but even if both participants in the channel sign a correct
> statement of truth, one of the participants can send funds out in the next
> second, invalidating that truth. While proof of ownership of on-chain UTXOs
> can be seen publicly in real time if they are spent, LN transactions aren't
> public like that. So any balance attestation is at best only valid the
> instant its taken, and can't be used as verification the money is still
> owned by the same channel partner in the next second.
>
>
> The same problem really also exists onchain --- a thief (or "thief") who
> has gotten a copy of the key can sign a transaction that spends it, one
> second after the proof-of-reserves is made.
>
> Really, though, the issue is that ownership of funds is conditional on
> *knowledge* of keys.
> And *knowledge* is easily copyable.
>
> Thus, it is possible that the funds that are "proven" to be the reserve of
> a custodian is actually *also* owned by someone else who has gotten to the
> privkeys (e.g. somebody threw a copy of it from a boating accident and a
> fearless scuba diver rescued it), and thus can also move the funds outside
> of the control of the custodian.
> This condition can remain for many months or years, as well, without
> knowledge of the custodian clients, *or* of the custodian itself.
>
> There is no way to prove that there is no alternate copy of the privkeys,
> hence "if only one could prove that he won't get into a boating accident".
>
> On the other hand, one could argue that at least the onchain proof
> requires more conditions to occur, so we might plausibly live with "we
> cannot prove we will never get into a boating accident but we can show
> evidence that we live in a landlocked city far from any lakes, seas, or
> rivers".
>
> Regards,
> ZmnSCPxj
>
>
>   a custodian Lightning node is unable to "freeze" a snapshot of its
> current state and make an atomic proof-of-reserves of *all* channels
>
>
> That would be a neat trick. But yeah, I don't know how that would be
> possible.
>
>
>   I believe it is one reason why custodian proof-of-reserves is not that
> popular ... it does not prove that the key will not get lost
>
>
> True, but at least if funds do get lost, it would be come clear far
> quicker. Today, an insolvent company could go many months without the
> public finding out.
>
>
> On Mon, Jul 5, 2021 at 5:09 PM ZmnSCPxj <ZmnSCPxj@protonmail.com> wrote:
>
>
> Good morning e,
>
>
> If only one could prove that he won’t get into a boating accident.
>
>
> At least in the context of Lightning channels, if one party in the channel
> loses its key in a boating accident, the other party (assuming it is a true
> separate person and not a sockpuppet) has every incentive to unilaterally
> close the channel, which reveals the exact amounts (though not necessarily
> who owns which).
>
> If the other party then uses its funds in a new proof-of-reserves, then
> obviously the other output of the unilateral close was the one lost in the
> boating accident.
>
>
> On the other hand, yes, custodians losing custodied funds in boating
> accidents is much too common.
>
> I believe it is one reason why custodian proof-of-reserves is not that
> popular --- it only proves that the funds were owned under a particular key
> at some snapshot of the past, it does not prove that the key will not get
> lost (or "lost and then salvaged by a scuba diver") later.
>
>
> Regards,
>
> ZmnSCPxj
>
>
>
> e
>
>
> On Jul 5, 2021, at 16:26, ZmnSCPxj via bitcoin-dev
> bitcoin-dev@lists.linuxfoundation.org wrote:
>
> Good morning Billy,
>
>
> I wonder if there would be some way to include the ability to prove
> balances held on the lightning network, but I suspect that isn't generally
> possible.
>
>
> Thinking about this in terms of economic logic:
>
> Every channel is anchored onchain, and that anchor (the funding txout) is
> proof of the existence, and size, of the channel.
>
> The two participants in the channel can sign a plaintext containing their
> node pubkeys and how much each owns.
>
> One of the participants should provably be the custodian.
>
>
> -   If the counterparty is a true third party, it has no incentive to lie
> about its money.
>
> -   Especially if the counterparty is another custodian who wants
> proof-of-reserves, it has every incentive to overreport, but then the first
> party will refuse to sign.
>
>      It has a disincentive to underreport, and would itself refuse to sign
> a dishonest report that assigns more funds to the first party.
>
>      The only case that would be acceptable to both custodians would be to
> honestly report their holdings in the Lightning channel.
>
>
> -   If the counterparty is a sockpuppet of the custodian, then the entire
> channel is owned by the custodian and it would be fairly dumb of he
> custodian to claim to have less funds than the entire channel.
>
>
> Perhaps a more practical problem is that Lightning channel states change
> fairly quickly, and there are possible race conditions, due to network
> latency (remember, both nodes need to sign, meaning both of them need to
> communicate with each other, thus hit by network latency and other race
> conditions) where a custodian Lightning node is unable to "freeze" a
> snapshot of its current state and make an atomic proof-of-reserves of all
> channels.
>
> Regards,
>
> ZmnSCPxj
>
>
> bitcoin-dev mailing list
>
> bitcoin-dev@lists.linuxfoundation.org
>
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 18254 bytes --]