26/27 already taken by #814 (at least, I proposed it move there to avoid clash)? Try 28/29?

Would it make sense to add optional auth_token field (up to 33 bytes?) so that one could only allow dual-funded channels for authorized peers? Or maybe to identify how much the other side wants to put into the channel.

Can be done by node ID too but this may lead to less UX hassle as the ID doesn't need to be transferred back-and-forth.

Yes, I mean perhaps it'd be beneficial to have a standardized field so that clients implement the interface of filling that in. Would be real shame if only LND could open authenticated channel to LND, Elair to Ecliar and c-lightning to c-lightning...

Basically have a way for the server to provide hex/base64 string that user fills in to get double-funded channel. This will be easier if all impls agree on the field being used.

I think this proposal is a great opportunity to provide a standard way of paying for liquidity (to help with the initial inbound liquidity problem). Maybe it's worth a separate PR that will enrich the interactive-tx protocol (and can be merged separately), I'll work on that once I've done a more thorough review of this PR.

The high-level idea is to simply introduce new TLVs to allow participants to:

• announce the fee they would charge to provide liquidity (funds on their side) -> probably in node_announcement
• request the remote participants to add X sats by paying that fee -> in the interactive-tx messages

Deducing the fee from the initial main outputs would then be deterministic and could be verified before finalizing the funding tx, which is great. It would remove the hackish uses of push_msat in trusted liquidity providers.

It would also provide an incentive for nodes to add funds to channels where they're not the initiator. Because without such a fee, it will be hard to automate the addition of funds when we're fundee, because we most likely don't want to add some of our funds to every channel that remote nodes open to us (otherwise they could easily deplete our utxo pool).

Any thought on this high-level proposal before I start diving into technical details?

I think this proposal is a great opportunity to provide a standard way of paying for liquidity (to help with the initial inbound liquidity problem). Maybe it's worth a separate PR that will enrich the interactive-tx protocol (and can be merged separately), I'll work on that once I've done a more thorough review of this PR.

The high-level idea is to simply introduce new TLVs to allow participants to:

• announce the fee they would charge to provide liquidity (funds on their side) -> probably in node_announcement
• request the remote participants to add X sats by paying that fee -> in the interactive-tx messages

Deducing the fee from the initial main outputs would then be deterministic and could be verified before finalizing the funding tx, which is great. It would remove the hackish uses of push_msat in trusted liquidity providers.

It would also provide an incentive for nodes to add funds to channels where they're not the initiator. Because without such a fee, it will be hard to automate the addition of funds when we're fundee, because we most likely don't want to add some of our funds to every channel that remote nodes open to us (otherwise they could easily deplete our utxo pool).

Any thought on this high-level proposal before I start diving into technical details?

I think this is a great idea! So great in fact that I've already written up a spec draft and have started working on implementing it. I just pushed up the draft version #878, and would love to collaborate, get your feedback on it.

Should we recommend to verify finalized interactive transaction against client's full-node transaction standardness verifier.

Hm. Ideally we would specify all of the "standardness" rules here so as not to be dependent on the 'standardness verifier' implementation, however that list seems to fairly changeable. Better to specify what to do in the case that the broadcast fails? (double spend the inputs/attempt another RBF).

Should we enrich failure modes ? ... In that case of scenario, honest double-spend of one's inputs after a timeout might be wiser.

Yeah, the 'double spending of an input to cancel an open' is a bit implicit here, it's probably worthwhile to call it out explicitly.

Should we introduce the interactive transaction section in its own BOLT ? As the 2 above points underscore it, multi-party funded transaction in the context of trust-minimized counterparty is far harder to get right than single-funded transaction. We migh land first this chunk, and from then enrich/harden the spec with best practices as we discover more edge cases and base layer support improves and we might add poodle or an equivalent solution to the privacy issues.

Yeah, I think this is a good idea, and something that we can do naturally/easily after this is merged (and more sections of the spec adopt it/use it)

Most recent push is whitespace only changes.

During yesterday's spec meeting, we discussed griefing attacks and pinning again, and suggested re-evaluating the option of making the opener send tx_signatures first regardless of who contributed more funds (see niftynei#9 for historical discussions).

There are three pinning vectors that apply to interactive-tx transactions:

1. Unconfirmed inputs that have a low ancestor feerate
2. Chaining low-feerate descendants to a change output of the transaction
3. Inputs for which the witness may be arbitrarily inflated to increase the weight of the transaction

Nodes can avoid the first one by setting the require_confirmed_inputs TLV.

The second one can be fixed by either constraining change outputs (to have a CSV 1 for example) or using v3 transactions when available. In practice though, as long as nodes ensure that the transaction has a reasonable feerate, it is very likely that it will confirm: it isn't that simple to reliably pin it.

The third one can be fixed by either sending tx_signatures last, or restricting what inputs may be used (e.g. to p2wpkh or key-path-only p2tr, or some template scripts). The main issue of changing the order of tx_signatures is that it breaks batching and adds a non-deterministic ordering rule to the protocol.

I'd like to highlight a few things to help guide the decision. First of all, if we keep the existing tx_signatures ordering, the attacker needs to lock more funds than the attacked node to be able to sign last. Also, this attack only works when the non-initiator decides to add funds to the dual-funded channel, which in practice will only happen if the initiator is paying for a liquidity ads, and the seller will apply some rate-limits based on who the buyer is (whether they have other channels in the network, if those channels are old, if the seller still has enough liquidity to spare, etc). When selling liquidity, you will need to score the buyer somehow to try to make the most yield, so that's simply another risk factor to take into account.

My personal opinion is that it's not worth introducing the tx_signatures ordering option, or at least not right now. It's unclear whether it will be necessary in practice, so I'd rather keep the protocol simpler, and implement policies on the liquidity ads seller side (restricting input types and who we sell liquidity to) if griefing becomes a widespread issue. If we later discover that we absolutely need to fix that, we can easily introduce this with a new feature bit. By that time, I hope that other solutions (mostly improvements to the bitcoin mempool policies) will give us better guarantees against pinning attacks as a whole.

@morehouse @niftynei what do you think?

During yesterday's spec meeting, we discussed griefing attacks and pinning again, and suggested re-evaluating the option of making the opener send tx_signatures first regardless of who contributed more funds (see niftynei#9 for historical discussions).

There are three pinning vectors that apply to interactive-tx transactions:

1. Unconfirmed inputs that have a low ancestor feerate

2. Chaining low-feerate descendants to a change output of the transaction

3. Inputs for which the witness may be arbitrarily inflated to increase the weight of the transaction

AFAICT, vector 2 only works in combination with vector 3.

There's also another category of pinning, where mempools get partitioned -- a low-feerate transaction that conflicts with the funding tx can be pinned in some mempools while the funding tx is in the victim's mempool.

Nodes can avoid the first one by setting the require_confirmed_inputs TLV.

Perhaps we make this the default instead of a TLV?

The second one can be fixed by either constraining change outputs (to have a CSV 1 for example) or using v3 transactions when available.

I'm not sure these are magic bullets either. Either one can require the victim to spend some extra fees to CPFP or double-spend the funding tx. Under sustained griefing, the victim's funds could be slowly burned as fees.

In practice though, as long as nodes ensure that the transaction has a reasonable feerate, it is very likely that it will confirm: it isn't that simple to reliably pin it.

Right, vector 3 is what enables the feerate to be reduced so that pinning can occur.

The third one can be fixed by either sending tx_signatures last, or restricting what inputs may be used (e.g. to p2wpkh or key-path-only p2tr, or some template scripts).

I think p2tr doesn't work, because even key-path spends can be inflated via the annex.

The main issue of changing the order of tx_signatures is that it breaks batching and adds a non-deterministic ordering rule to the protocol.

I'd like to highlight a few things to help guide the decision. First of all, if we keep the existing tx_signatures ordering, the attacker needs to lock more funds than the attacked node to be able to sign last.

Sort of, but those funds don't need to actually be the attacker's. If Mallory targets 2 (or more) LSPs at once, she can use the LSPs' inputs against each other, ensuring she sends tx_signatures last. Mallory only needs to lock enough funds to pay the liquidity ads.

Also, this attack only works when the non-initiator decides to add funds to the dual-funded channel, which in practice will only happen if the initiator is paying for a liquidity ads, and the seller will apply some rate-limits based on who the buyer is (whether they have other channels in the network, if those channels are old, if the seller still has enough liquidity to spare, etc). When selling liquidity, you will need to score the buyer somehow to try to make the most yield, so that's simply another risk factor to take into account.

My personal opinion is that it's not worth introducing the tx_signatures ordering option, or at least not right now. It's unclear whether it will be necessary in practice, so I'd rather keep the protocol simpler, and implement policies on the liquidity ads seller side (restricting input types and who we sell liquidity to) if griefing becomes a widespread issue. If we later discover that we absolutely need to fix that, we can easily introduce this with a new feature bit. By that time, I hope that other solutions (mostly improvements to the bitcoin mempool policies) will give us better guarantees against pinning attacks as a whole.

Yeah, I'm not sure either. I made a Discourse post discussing liquidity griefing in depth. It would be great to get some discussion going over there.

Sort of, but those funds don't need to actually be the attacker's. If Mallory targets 2 (or more) LSPs at once, she can use the LSPs' inputs against each other, ensuring she sends tx_signatures last. Mallory only needs to lock enough funds to pay the liquidity ads.

Right, that's a good point!

Perhaps we make this the default instead of a TLV?

In the longer term, we should be able to get rid of this restriction, so I'd rather keep it opt-in. We discussed this a while ago, there are scenarios where it makes sense to use unconfirmed inputs to optimize liquidity usage (high fees and chains of unconfirmed txs), and if the other party is honest, this doesn't create any issue.

I'm not sure these are magic bullets either. Either one can require the victim to spend some extra fees to CPFP or double-spend the funding tx. Under sustained griefing, the victim's funds could be slowly burned as fees.

True, this can create a slow drain on the victim's funds if the attack is sustained. But that's only under the assumption that the transactions never confirm on their own, which they sometimes should if the feerate is reasonable, in which case it costs fees for the attacker, not the victim. If the feerate was too low and we're using v3, then a double-spend isn't necessarily costing anything to the victim, that double-spend would just pay the expected current feerate (and if this is double-spent in another liquidity ads attempt, it won't be the seller who pays the on-chain fees).

In the longer term, we should be able to get rid of this restriction, so I'd rather keep it opt-in.

How will we do this?

I'm not sure these are magic bullets either. Either one can require the victim to spend some extra fees to CPFP or double-spend the funding tx. Under sustained griefing, the victim's funds could be slowly burned as fees.

True, this can create a slow drain on the victim's funds if the attack is sustained. But that's only under the assumption that the transactions never confirm on their own, which they sometimes should if the feerate is reasonable, in which case it costs fees for the attacker, not the victim. If the feerate was too low and we're using v3, then a double-spend isn't necessarily costing anything to the victim, that double-spend would just pay the expected current feerate (and if this is double-spent in another liquidity ads attempt, it won't be the seller who pays the on-chain fees).

Good points. I did a little math to see what a typical griefing might cost an LSP.

Some recent mempool feerates:

• Minimum relay: 10 sat/vB
• 1 week confirmation: 20 sat/vB
• 6 hour confirmation: 80 sat/vB

Let's say we use 80 sat/vB for a funding transaction and the LSP's portion of the transaction is 250 vB, so they allocate 80 * 250 = 20,000 sat for the transaction fee. Let's say the total transaction is 500 vB, so the total fee for the transaction is 40,000 sat.

The attacker inflates their witnesses to reduce the feerate to 10 sat/vB and also chains a 1,000 vB descendant transaction (the max allowed by v3 policy) with a fee of 10,000 sat (10 sat/vB). This package is unlikely to confirm for a very long time.

If the LSP wishes to double-spend their griefed UTXOs, they must now spend 40,000 + 10,000 = 50,000 sat. If they use another 500 vB dual-funded transaction to double spend, the required feerate for that transaction would be 50,000 sat / 500 vB = 100 sat/vB This is 25% more than the expected current feerate.

Suppose further that the attacker is doing repeated griefing and this new dual-funded transaction is actually with another node controlled by the attacker. The attacker can grief this transaction as well, causing the double-spend fee to increase to 50,000 + 10,000 = 60,000 sat (120 sat/vB). And so on.

How will we do this?

If an attacker lowers the effective feerate by using low-feerate ancestors, or double-spends one of the ancestor transactions, this is free to double-spend: a competing transaction can use the same feerate without any additional cost. The only cumbersome part is how long you wait before figuring out that something is fishy with that transaction and you should unlock its utxos. For that, you need a heuristic that "this tx seems to have a good enough feerate, but it hasn't confirmed for X blocks, so let's mark it as malicious".

The attack becomes more painful when coupled with a descendant chain, and IMO the only way to fix this that doesn't have annoying drawbacks is to fix mempool policies ™️ 🤞 (v3 txs / cluster mempool).

If we don't need to care about descendant pinning, this is thus not so much of an issue and confirmed inputs shouldn't be the default (as they create inefficient liquidity usage). I've detailed that point a bit more on your delving bitcoin post, maybe we should keep the discussion over there where it will be accessible to more readers than on this PR?

Good points. I did a little math to see what a typical griefing might cost an LSP.

Nice work! But this pinning attack does not reliably work though, so I believe it's not the end of the world. To be able to effectively lower the feerate of the funding transaction, the attacker needs to have low-feerate ancestors. But since those ancestors are low-feerate, they will be dropped from a fraction of the network's mempools, which will allow an honest transaction at the normal feerate to go through (because in those mempools, that honest transaction isn't a double-spend, the conflicting transactions have been evicted). That's what @TheBlueMatt mentioned a few times, that in practice today, based on his experiments, even low-ish feerate transactions are still able to propagate and confirm.

Since this PR hasn't been merged yet, I'd like to get everyone's feedback on adding yet another change before integrating this: niftynei#18

After having hands-on experience doing support for users who are affected by this, I think it would be best to immediately provide the tools to get rid of this issue entirely now that taproot is a thing. The PR linked contains two design options to achieve that, let me know what you think is best. Even though we have running code using the current messages, I'm fully open to making backwards-incompatible changes to the spec if necessary to ensure a cleaner protocol in the future without the need for additional feature bits/negotiation.

@niftynei @morehouse @dunxen @ddustin

After figuring out the attack vectors in more details, I've retracted niftynei#18 (more details directly on that PR).

So I'm stilli ACKing 27ffef4 and I think we should consider merging this PR!

As discussed in yesterday's spec meeting, this PR has widespread ACKs so we can merge it whenever you want @niftynei 🎉 🎉

The only remaining open comment is this one: #851 (comment), up to you to decide what you want to do here (and we can also re-work the wording or rationale in a follow-up PR if necessary).

Thanks @t-bast for shepherding this through for the last few months (years?). Would not have happened without his tireless focus on getting this done!

Now we splice?? :)