Simple lock/unlock mechanism

With the recent trend of physically robbing people for bitcoin (or other cryptocurrencies), I thought it would be beneficial to introduce a standard for locking up a portion of your bitcoin in a simple 'gotta-wait-awhile' system.

The idea is to simply create a transaction spending a set of UTXOs to a P2SH address with an OP_CSV attached to it, and then throw away the private keys.

To spend the bitcoin, you would have to broadcast the transaction and wait the given amount of time, then use the new txout.

There are several ways to shoot yourself in the foot trying to do this manually, but e.g. Bitcoin Core could handle this in a fairly straightforward manner by introducing two new commands, which I call freeze and unfreeze:

bitcoin-cli freeze [amount OR array of txid+vout objects] [days, default 1]

E.g. bitcoin-cli freeze 10 5 E.g. bitcoin-cli freeze ["abc123:1", "def346:0"] 3

The unfreeze command would by default list all freezes:

bitcoin-cli unfreeze txid days status bcd123 5 frozen dca999 3 frozen

including the txid would broadcast the unfreeze and status would become 'thawing' until the amount becomes available:

bitcoin-cli unfreeze bcd123

bitcoin-cli unfreeze txid days status bcd123 5 thawing dca999 3 frozen

The benefit of this is that it becomes physically impossible for you to spend the coins until you thaw them, and if this becomes standard enough, it should disincentivize potential robbers as it would be expected that you keep most of your assets locked up. They could of course hold you hostage until the period is over, which may be worse, but I think that kind of operation would be substantially more difficult than a simply rob-and-run.

The drawback is that you have to broadcast a transaction in order to spend the content, and you cannot bump the fee so the transaction could get stuck in a high-fee situation.

-Kalle.

A few p.s.'es:

1. Graftroot probably breaks this (someone could just sign the time-locked output with a script that has no time-lock).

2. Address reuse of discarded privkeys would be a problem. A way to alleviate might be that freezing includes a send to a new address and the privkeys for that new one are discarded instead. One extra transaction, but reduces the risk of accidentally throwing away `donations4mybook` vanity keys.

-Kalle.

On Wed, Feb 28, 2018 at 3:47 AM, アルム　カールヨハン <karl@dglab•com> wrote:

> With the recent trend of physically robbing people for bitcoin (or
> other cryptocurrencies), I thought it would be beneficial to introduce
> a standard for locking up a portion of your bitcoin in a simple
> 'gotta-wait-awhile' system.
>
> The idea is to simply create a transaction spending a set of UTXOs to
> a P2SH address with an OP_CSV attached to it, and then throw away the
> private keys.
>
> To spend the bitcoin, you would have to broadcast the transaction and
> wait the given amount of time, then use the new txout.
>
> There are several ways to shoot yourself in the foot trying to do this
> manually, but e.g. Bitcoin Core could handle this in a fairly
> straightforward manner by introducing two new commands, which I call
> freeze and unfreeze:
>
> bitcoin-cli freeze [amount OR array of txid+vout objects] [days, default 1]
>
> E.g. bitcoin-cli freeze 10 5
> E.g. bitcoin-cli freeze ["abc123:1", "def346:0"] 3
>
> The unfreeze command would by default list all freezes:
>
> bitcoin-cli unfreeze
> txid         days     status
> bcd123   5           frozen
> dca999   3           frozen
>
> including the txid would broadcast the unfreeze and status would
> become 'thawing' until the amount becomes available:
>
> bitcoin-cli unfreeze bcd123
>
> bitcoin-cli unfreeze
> txid         days     status
> bcd123   5           thawing
> dca999   3           frozen
>
> The benefit of this is that it becomes physically impossible for you
> to spend the coins until you thaw them, and if this becomes standard
> enough, it should disincentivize potential robbers as it would be
> expected that you keep most of your assets locked up. They could of
> course hold you hostage until the period is over, which may be worse,
> but I think that kind of operation would be substantially more
> difficult than a simply rob-and-run.
>
> The drawback is that you have to broadcast a transaction in order to
> spend the content, and you cannot bump the fee so the transaction
> could get stuck in a high-fee situation.
>
> -Kalle.

On Wed, Feb 28, 2018 at 04:34:18AM +0000, アルム カールヨハン via bitcoin-dev wrote:

> 1. Graftroot probably breaks this (someone could just sign the
> time-locked output with a script that has no time-lock).

Making the graftroot key be a 2-of-2 muSig with an independent third party that commits to only signing CLTV scripts could avoid this. Making it 3-of-3 or 5-of-5 could be even better if you can find multiple independent services that will do it.

Cheers, aj

Coincidentally I had thought of something similar to what Kalle posted about a kind of software only time-lock vault, and described the idea to a few people off-list. Re. Root incompatibility, if the key is deleted (as it must be) then a delegated signature can not be made that bypasses the CSV timeout restriction, so Root should not be incompatible with this. I think it would be disadvantageous to mark keys as Rootable vs not in a sighash sense, because then that is another privacy/fungibility loss eroding the uniformity advantage of Root when the delegate is not used.

One drawback is deleting keys may itself be a bit difficult to assure with HD wallet seeds setup-time backup model.

As Anthony described I think, a simpler though less robust model would be to have a third party refuse to co-sign until a pre-arranged time, and this would have the advantage of not requiring two on-chain transactions.

With bulletproofs and CT rangeproofs / general ECDL ZKPS there is the possibility to prove things about the private key, or hidden attributes of a public key in zero-knowledge. Kind of what we want is to place private key covenants, where we have to prove that they are met without disclosing them. For example there is a hidden CSV and it is met OR there is no hidden CSV so it is not applicable.

Adam

On 28 February 2018 at 23:30, Anthony Towns via bitcoin-dev <bitcoin-dev@lists•linuxfoundation•org> wrote:

> On Wed, Feb 28, 2018 at 04:34:18AM +0000, アルム カールヨハン via bitcoin-dev wrote:
>> 1. Graftroot probably breaks this (someone could just sign the
>> time-locked output with a script that has no time-lock).
>
> Making the graftroot key be a 2-of-2 muSig with an independent third party
> that commits to only signing CLTV scripts could avoid this. Making it
> 3-of-3 or 5-of-5 could be even better if you can find multiple independent
> services that will do it.
>
> Cheers,
> aj
>
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev@lists•linuxfoundation•org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev

On Wed, Feb 28, 2018 at 10:30 PM, Anthony Towns <aj@erisian•com•au> wrote:

> On Wed, Feb 28, 2018 at 04:34:18AM +0000, アルム カールヨハン via bitcoin-dev wrote:
>> 1. Graftroot probably breaks this (someone could just sign the
>> time-locked output with a script that has no time-lock).
>
> Making the graftroot key be a 2-of-2 muSig with an independent third party
> that commits to only signing CLTV scripts could avoid this. Making it
> 3-of-3 or 5-of-5 could be even better if you can find multiple independent
> services that will do it.

That kind of defeats the purpose. If you go through the trouble of doing that, you can just do multisig and skip the freezing part entirely. A robber would have to get you and the cosigner to sign in both cases, and the CLTV could be overridden with graftroot.

On Wed, Feb 28, 2018 at 11:36 PM, Adam Back <adam.back@gmail•com> wrote:

> Coincidentally I had thought of something similar to what Kalle posted
> about a kind of software only time-lock vault, and described the idea
> to a few people off-list.  Re. Root incompatibility, if the key is
> deleted (as it must be) then a delegated signature can not be made
> that bypasses the CSV timeout restriction, so Root should not be
> incompatible with this.  I think it would be disadvantageous to mark
> keys as Rootable vs not in a sighash sense, because then that is
> another privacy/fungibility loss eroding  the uniformity advantage of
> Root when the delegate is not used.

1. Create TX1=(tx, sig) from UTXO A to p2sh B which has a CSV timelock. Discard privkey A. 2. After broadcasting TX1, you need privkey B to spend it. 3. Use graftroot and privkey B with a script without timelock to spend B.

The robber can simply force you to execute step 3, since you have the privkey to B.

> One drawback is deleting keys may itself be a bit difficult to assure
> with HD wallet seeds setup-time backup model.

That's a good point. Even more of a reason to include as part of 'freezing' a send to a new ephemeral key as 'initialization'. Sucks to pay triple fees though (freeze ephemeral + unfreeze + actual use).

> As Anthony described I think, a simpler though less robust model would
> be to have a third party refuse to co-sign until a pre-arranged time,
> and this would have the advantage of not requiring two on-chain
> transactions.

I was hoping there was a way for a person to simply lock-up the major portion of their coins easily.

As a sidenote: a security firm (e.g. one that comes to your house when the alarm goes off) could have a service where seeing an unfreeze transaction which you have told them about without you giving a heads up beforehand is equal to alarm going off.

-Kalle.

On Thu, Mar 1, 2018 at 5:11 AM, アルム　カールヨハン <karl@dglab•com> wrote:

> That kind of defeats the purpose. If you go through the trouble of
> doing that, you can just do multisig and skip the freezing part
> entirely. A robber would have to get you and the cosigner to sign in
> both cases, and the CLTV could be overridden with graftroot.

I think I'm confused on this. To use graftroot it has to be a pubkey, not a p2sh thing.

-Kalle.