This writeup is of interest to node operators and Lightning Network wallet end-users, even those without programming experience.
Inbound liquidity is, somehow, always a concern for the Lightning Network. It is trivial for a new node to get outbound liquidity by simply allocating some onchain funds to new channels to the rest of the network. But for some reason, it is apparently a difficult task to acquire inbound liquidity.
Or in other words, it is easy for a new LN wallet to send funds, but hard for a new LN wallet to receive funds. Somehow, an LN wallet has to get this thing called “inbound liquidity” in order to actually receive funds.
The market around LN has come up with a few solutions to this problem:
- Custodial wallets. An expert node operator contacts other, existing nodes to request incoming liquidity from them, or otherwise acquire incoming liquidity on their own node. Then that operator sells the LN wallet service (both outbound and inbound liquidity) of their node to end-users, in exchange for knowledge of all their LN payments, and the irrevocable and very transferable right to steal all their LN funds at any time.
- LSPs offering services to make channels to new LN wallets and give them incoming liquidity. This service can be paid for explicitly by charging a direct fee for the incoming liquidity, or implicitly by requiring an unpublished (often called oxymoronically as “private”) channel (thus getting knowledge of all LN sends and receives via that channel, as per axiom of terminus), or again implicitly by charging usurious offchain feerates for use of the channel.
The above solutions usually have negative implications for the privacy and security of your LN funds and fund transfers.
More recently (mid 2022), Lisa Neigut has been designing dual-funding, also known as “openchannelv2”. This largely makes the LSP-solution above much easier. The dual-funding protocol also includes the liquidity ads sub-proposal, which further facilitates the above LSP-solution. Liquidity ads encourages the explicit charge for a published channel rather than implicit fees from reselling information extracted from unpublished (laughably known as “private”) channels, or charging usurious feerates.
Fundamentally speaking, we can consider both approaches as “selling inbound liquidity”. The sellers of the inbound liquidity are able to acquire inbound liquidity for their own nodes, and then let you purchase the use of that inbound liquidity. Typically, this inbound liquidity is acquired wholesale by direct dealings with operators of large published nodes, often operated or owned by existing Bitcoin whales. The inbound liquidity is then resold at retail to smaller individual end-users.
However, it is a better proposition for the inbound liquidity to be sold for a monetary fee, then to pay using your information. Information about you can be used to manipulate and control you in order to extract tribute from you later. Thus, solutions that involve end users paying for incoming liquidity by sacrificing their information should be avoided.
Non-Custodiality And Privacy
TLDR: non-custodial wallets that always connect to a fixed LSP (or set of LSPs) are still no good in terms of keeping your information safe, though as non-custodial they do keep your actual funds safe. The rest of this subsection is deeply technical.
Some developers of non-custodial LN wallets have considered setting up their own LSP, and having their non-custodial LN wallets always connect to that LSP. The idea is that the developers also get their inbound liquidity wholesale and then provide it “for free” to end-users, while using non-custodial techniques so that their node never holds end-user funds.
However, if your non-custodial wallet always connects to one or more LSP(s) run by a single entity, then you are paying for its services using your information: the LSP knows that all transfers on your channels with it are your activity, because your non-custodial wallet cannot connect to anyone else. (If the wallet can in theory connect to “your own node!” then just run your own published node and use a mobile remote control UI to that published node, instead of controlling your own node by tying up more Bitcoins into a channel from your mobile non-custodial wallet, which you own, to your own node, which you also own, so all transfers between them are of null utility since you own both nodes anyway, so just use a remote control instead and avoid tying up more funds.)
No amount of promising from these developers can truly protect your information. In particular, with the LN as of mid 2022, it is impossible to delete the information even if the operators really really really want to: Poon-Dryja channel operators MUST keep a copy of every HTLC that ever existed on a channel, otherwise old state could have an old HTLC that cannot be retrieved since the hash and timelock details are not known. The HTLC data can only be safely deleted if the channel is closed (at which point you lose access to the LN). You might trust them now, but their values may change in the future, they or their employer may be bought out by some entity with very different values, their LSP can be hacked and the information stolen, etc.
In theory the non-custodial LN wallet developers could release the code for their LSP so that anyone else can run an LSP as well. While better than fixing your non-custodial wallet to use LSP(s) from a single entity, we should compare this to the situation with Electrum servers. Electrum servers do not require that the server operators lock up funds for the Electrum server, so it is far easier for a random nameless volunteer to run one, whereas LSPs should be large published nodes or else their capacity may be overwhelmed. LSPs need to somehow acquire the inbound liquidity wholesale (again, often by directly dealing with other large published nodes) that they will provide to the end-users. It is certainly possible to do so, but the barriers are higher than for Electrum servers, and the undeletable log data is still a problem.
Inbound Liquidity Via Offchain-to-Onchain Swaps
During the blocksize debate, one of the points against the Lightning Network, and therefore the smallblock strategy, was this:
Largeblocker: LN channels start in one direction only and you cannot receive by it! That sucks!
Smallblocker: Well, if you send out funds first, then you get inbound liquidity so you can then receive afterwards.
Largeblocker: So I have to spend first before I can receive?? Nice UX, LN! What if I am not interested in spending my Bitcoins, I just want to buy more Bitcoins so I have to receive it?
However, an important key insight was this:
Smallblocker: 1 BTC = 1 BTC. You can spend Bitcoins on the Lightning Network to buy an equivalent amount of Bitcoins on the blockchain, which gets you inbound liquidity (due to sending out the funds over the channel), while still keeping your Bitcoins as Bitcoins.
Basically, you use an offchain-to-onchain swap service, such as Boltz Exchange or Lightning Loop, and because you send out funds over the LN, you get inbound liquidity. At the same time, you can then take the onchain Bitcoins and lock them into new channels, to return the same amount into a readily-spendable outbound liquidity.
This has several advantages:
- If you contact the swap service over Tor, and mix the received onchain coins using CoinJoin or CoinSwap, then the swap service cannot identify which node on the network is using their service to get inbound liquidity, thus cannot resell your information.
- You are not tied to specific LSPs, and you can connect channels to any published node on the network. This improves your decentralization and resilience, as there are far more published nodes than there are LSPs that sell inbound liquidity.
- You are distant from the swap service you are buying the inbound liquidity from, unlike the case of an LSP which must be your peer. The swap service is effectively transferring their inbound liquidity to you, not leasing their inbound liquidity (like in the LSP case). Once you have purchased the liquidity, the liquidity of the swap service no longer matters to you, and even if the swap servie manages to lose all inbound liquidity, you can still receive since you are distant from the swap service. Compare this to the LSP case, where if the LSP manages to lose inbound liquidity, even if you have channels with the LSP where the LSP can send to you, if the LSP itself cannot receive from the rest of the network, it cannot forward to you and you cannot receive via that LSP since it is your broker to the rest of the network.
- You can do this manually using e.g. the Boltz Exchange website, even if your LN wallet does not support it directly.
Strangely, this technique seems to be underutilized. To my knowledge (mid 2022), only the Electrum LN wallet has direct support for this technique, and provides a convenient button on its UI to perform a swap for the purposes of getting inbound liquidity. CLBOSS also uses this technique to buy inbound liquidity, and does so automatically when it judges that the node under management has insufficient inbound liquidity. People are somehow amazed at how CLBOSS can “magically” get balanced channels with good inbound and outbound liquidity, yet this technique can be done perfectly manually and is very simple, both in concept and execution. CLBOSS just adds automation to do this technique without user intervention.
The technique is not perfect.
You need Bitcoins to first create channels, then reverse the
direction of the channels via this offchain-to-onchain swap
technique, then create new channels to put your onchain
funds back offchain.
Typically, wallet developers wnat to demo their wallet by
first sending to prospective end-users some nominal amount
of funds, then show that it can be spent.
But this case can be handled by 0-conf
channels, and once you have a channel, you can reverse its
direction using this technique.
Swap Service Point-of-View
We might wonder about the point-of-view of the swap service.
As far as the swap service knows, it is selling onchain Bitcoins for offchain Bitcoins.
However, we should note that in order to receive offchain Bitcoins over Lightning, the swap service itself has to have its own inbound liquidity.
When the inbound liquidity is then subsequently used up during the buying of onchain Bitcoins, then the swap service has to somehow acquire new inbound liquidity and new onchain Bitcoins.
How the service does so, will depend on the operators of the service. The swap service may have deals with other large LN node operators to get inbound liquidity, and deals with miners to get onchain Bitcoins.
Basically, this operation transfers the inbound liquidity of the swap service over to the inbound liquidity of the user of this technique. The surcharge the swap service imposes then is the fee for the purchase of this inbound liquidity. Like in the LSP case, the node acquires inbound liquidity wholesale, and then sells it at retail to end-users.
A difference between the swap service and the LSP case, as noted above, is that the LSP is really renting out its inbound liquidity to its clients. For example, if the LSP has 1 BTC in inbound liquidity to the rest of the network, it could offer 1 BTC of liquidity to one client and also 1 BTC to another client. It is gambling that the clients will not simultaneously require the inbound liquidity, because if both of them try to use the full, sold capacity of 1 BTC, then the actual lack of liquidity of the LSP will prevent them from receiving.
With the swap service instead, the swap service cannot sell inbound liquidity it does not have. If the swap service has only 1 BTC of inbound liquidity, then if one client purchases that 1 BTC of inbound liquidity (by paying it for onchain Bitcoins, which converts the client outbound liquidity to inbound liquidity and the swap service inbound liquidity to outbound liquidity) then the swap service cannot sell the same 1 BTC of inbound liquidity to another client later, since it no longer has any inbound liquidity to sell.
With the use of a swap service rather than an LSP, the clients do not need to audit the inbound liquidity of the swap service, whereas with the LSP, the clients need to monitor and audit the inbound liquidity of the LSP. Unfortunately, the liquidity of the LSP represents private information of the LSP, and it is unsafe for the LSP to share this information.
1) LSP sells 2 BTC of inbound liquidity, despite only having 1 BTC of inbound liquidity. 2) Client 1 consumes its 1 BTC of inbound liquidity, which also consumes the 1 BTC of inbound liquidity of the LSP. Client 2 thinks it still has 1 BTC of inbound liquidity, but actually cannot receive except from Client 1.
1) Client 1 and Client 2 connect to the Lightning Network, where Swap is offerring an offchain-to-onchain swap service. 2) Client 1 outright purchases the available inbound liquidity from Swap. Client 2 is unable to purchase the inbound liquidity from Swap, but knows it has no inbound liquidity. Client 2 can choose to use a different Swap service, without having to close the existing channel it already made.
CLBOSS Swap Decisionmaking
CLBOSS is an automated manager, so it has to have some policy on when to trigger a swap. In CLBOSS as of this writing (mid 2022) there are two conditions that would cause an offchain-to-onchain swap.
Boss::Mod::NodeBalanceSwapper: triggers by comparing the inbound liquidity versus the total of the inbound and outbound liquidity. If the inbound liquidity is less than 30% of the total, and onchain fees are low, it triggers a swap. If the inbound liquidity is less than 2.5% of the total, it triggers a swap even if onchain fees are high. This always sends out 2/3 of the difference between outbound and inbound liquidity.
Boss::Mod::NeedsOnchainFundsSwapper: triggers when a small amount of onchain funds is detected, but the amount is too small for a feasible practical channel. This will typically occur if for example a channel needs to get closed, and only a small amount of money happens to be on our side of the channel-to-close. Rather than HODL this money onchain, CLBOSS will try to make an offchain-to-onchain swap to get more onchain funds, and form a new, larger channel. This only triggers if onchain fees are low.
The main policy that buys inbound liquidity is the
effectively, preemptively buy inbound liquidity to
match a planned outbound liquidity when we later build a
new channel to the network.
The swap system of CLBOSS has sufficient abstraction that it can use any number of protocols to talk with the swap service. It also has the ability to use any number of actual swap services.
However, as of this writing (mid 2022), only one protocol is used, the Boltz protocol. This is because the Boltz API is well-documented.
It would be nice to also support using the Lightning Labs Lightning Loop. However, it has the disadvantage that the client-server API is undocumented other than the client Go code, and there is no open-source implementation of the server, only the client. Boltz has released an open-source server code as well, so in theory anyone can start up a new Boltz server.
As noted, CLBOSS can use multiple swap services of various protocols. In particular, the open-source Boltz server means that in theory multiple servers, run by different entities, can arise which use the same protocol. The CLBOSS code includes two swap services that both use the Boltz protocol: the Boltz server itself, and one run over a hidden service by AutonomousOrganization. However, the latter service appears to be moribund, and may be removed at some point in the future.
I encourage others to deploy new swap services, rather than LSPs.
CLBOSS immediately sends onchain funds directly to the node under management. This means that the swap service can learn who is using its services by simply observing the blockchain and seeing the funds it sent out onchain being spent into channels of a specific node. At some point in the future, it would be nice to add the option to use some kind of CoinJoin or CoinSwap service to separate the funds from the swap service, and prevent the swap service from reselling that information. In particular, an LSP has to know your node in order to give you inbound liquidity, while at least with a swap service, there is a way for you to hide your node from the swap service, it just has not been implemented on CLBOSS yet (mid 2022).
Even with the lack of privacy against swap service in practice (since CLBOSS does not yet use an anonymizing layer to protect your identity from the swap service), there is still a slight advantage. The swap service will learn your node after-the-fact, when it has already given you the inbound liquidity, and once it has learned your node, cannot revoke that inbound liquidity it has given to you. The LSP has to know your node before-the-fact and can deny you the inbound liquidity in the first place, and can close the channel after-the-fact if it changes its mind about you.
Addendum: “Private” Channels Are Not Private: Unpublished Channels Delenda Est
Commonly referred to by the misnomer “private channel”, unpublished channels leak your every offchain send and receive to your counterparty.
The reason for this is simple: only those who know the channel can use it for their payments.
If your channel is unpublished, then only two people know the channel: you and your LSP.
If the channel has activity, and the LSP is not the one that triggered that activity, then by process of elimination, it knows that you are the one using the channel for that activity. After all, only the two of you know the channel. In that case, it can assign with 100% probability those activity to you, and resell that information to others.
If you value your privacy and your right to not have your information resold, then use published channels.
Now you might quibble and say “yes but I could use the unpublished channel in an invoice so the channel is now known by another entity”. But if you are going to rely on that, just publish the channel outright in the first place. That increases your anonymity set to “all LN users”.
The axiom of terminus is just the observation that an unpublished channel can only be used at the start or the end of a payment route, and not at any intermediate hop. Thus, any observer along the route who sees a multihop payment route from or to an unpublished channel knows immediately that the route starts or stops at the node at the other end. This is just a reformulation of the above observation, that an unpublished channel is known only by two entities and if one entity sees activity and knows it is not using the channel, obviously the activity was from the other entity.