This week’s newsletter describes the block parsing bug affecting BTCD and LND last week, summarizes discussion about a planned Bitcoin Core feature change related to replace by fee, outlines research about validity rollups on Bitcoin, shares an announcement about a vulnerability in the draft BIP for MuSig2, examines a proposal to reduce the minimum size of an unconfirmed transaction that Bitcoin Core will relay, and links to an update of the BIP324 proposal for a version 2 encrypted transport protocol for Bitcoin. Also included are our regular sections with summaries of changes to services and client software, announcements of new releases and release candidates, and descriptions of notable merges to popular Bitcoin infrastructure projects.

News

  • Block parsing bug affecting BTCD and LND: on October 9th, a user created a transaction using taproot with a witness containing nearly a thousand signatures. The consensus rules for taproot don’t place any direct limits on the size of witness data. This was a design element discussed during taproot’s development (see Newsletter #65).

    Shortly after the large-witness transaction was confirmed, users began to report that the BTCD full node implementation and LND Lightning Network implementation were failing to provide data from the most recent blocks that were available to Bitcoin Core full nodes. For BTCD nodes, this meant that transactions which had been recently confirmed were being reported as still unconfirmed. For LND, it meant that new channels that had recently become ready to use weren’t being reported as fully open.

    A developer for both BTCD and LND fixed the problem in BTCD’s code, which LND uses as a library, and quickly released new versions for both LND (as mentioned in last week’s newsletter) and BTCD. All users of BTCD and LND should upgrade.

    Until a user upgrades their software, they will suffer the lack-of-confirmation problems described above and may also be vulnerable to several attacks. Some of those attacks require access to significant hash rate (making them expensive and, hopefully, impractical in this case). Other attacks, particularly those against LND users, require the attacker to risk losing some of their funds in a channel, which is also hopefully a sufficient deterrent. We again recommend upgrade and, further, we recommend that anyone using any Bitcoin software sign up for security announcements from that software’s development team.

    After the above disclosures, Loki Verloren posted to the Bitcoin-Dev mailing list to suggest that direct limits be added to taproot’s witness size. Greg Sanders replied to note that adding limits now would not only increase code complexity but could also lead to people losing their money if they already received bitcoins to a script which requires a large witness to spend.

  • Transaction replacement option: as reported in Newsletters #205 and #208, Bitcoin Core merged support for a mempoolfullrbf configuration option which defaults to the existing Bitcoin Core behavior of only allowing RBF replacement of transactions containing the BIP125 signal. However, if a user sets the new option to true, their node will accept and relay replacements for transactions that don’t contain the BIP125 signal, provided the replacement transactions follow all of Bitcoin Core’s other rules for replacements.

    Dario Sneidermanis posted to the Bitcoin-Dev mailing list that this new option may create problems for services which currently accept unconfirmed transactions as final. Although it’s been possible for years for users to run non-Bitcoin Core software (or patched versions of Bitcoin Core) that allow unsignaled full1 transaction replacement, there’s no evidence that software is widely used. Sneidermanis believes an easily accessible option in Bitcoin Core might change that by allowing enough users and miners to enable full RBF and make unsignaled replacement reliable. More reliable unsignaled replacement would also make it more reliable to steal from services that accept unconfirmed transactions as final, requiring those services to change their behavior.

    In addition to describing the problem and providing a detailed description of how services choose when to accept unconfirmed transactions, Sneidermanis also proposed an alternative approach: remove the configuration option from the upcoming Bitcoin Core release but also add code that will enable full RBF by default at a future moment. Anthony Towns posted several options for consideration and opened a pull request that implements a slightly modified version of Sneidermanis’s proposal. If merged and released in its current state, Towns’s PR will enable full RBF by default starting 1 May 2023. Users objecting to full RBF will still be able to prevent their nodes from participating by setting the mempoolfullrbf option to false.

  • Validity rollups research: John Light posted to the Bitcoin-Dev mailing list a link to a detailed research report he prepared about validity rollups—a type of sidechain where the current sidechain state is compactly stored on the mainchain. A user of the sidechain can use the state stored on the mainchain to prove how many sidechain bitcoins they control. By submitting a mainchain transaction with a validity proof, they can withdraw bitcoins they own from the sidechain even if the operators or miners of the sidechain try to prevent the withdrawal.

    Light’s research describes validity rollups in depth, looks at how support for them could be added to Bitcoin, and examines various concerns with their implementation.

  • MuSig2 security vulnerability: Jonas Nick posted to the Bitcoin-Dev mailing list about a vulnerability he and several others discovered in the MuSig2 algorithm as documented in a draft BIP. In short, the protocol is vulnerable if an attacker knows a user’s public key, a tweak to that public key that the user will sign for (such as with BIP32 extended pubkeys), and can manipulate which version of the key the user will sign for.

    Jonas Nick believes the vulnerability “should only apply in relatively rare cases” and encourages anyone using (or soon planning to use) MuSig2 to reach out to him and his co-authors with questions. The draft BIP for MuSig2 is expected to be updated soon to address the issue.

  • Minimum relayable transaction size: Greg Sanders posted to the Bitcoin-Dev mailing list a request for Bitcoin Core to relax a policy added to make it harder to exploit the CVE-2017-12842 vulnerability. This vulnerability allows an attacker who can get a specially-crafted 64 byte transaction confirmed into a block to trick lightweight clients into believing one or more different arbitrary transactions were confirmed. E.g., innocent user Bob’s Simplified Payment Verification (SPV) wallet might display that he’d received a million BTC payment with dozens of confirmations even though no such payment was ever confirmed.

    When the vulnerability was only privately known among a few developers, a limit was added to Bitcoin Core preventing relay of any transaction with fewer than 85 bytes (not counting witness bytes), which is about the smallest size that can be created using standard transaction templates. This would require an attacker to get their transaction mined by software not based on Bitcoin Core. Later, the consensus cleanup soft fork proposal suggested permanently fixing the problem by disallowing any transactions less than 65 bytes in size from being included in new blocks.

    Sanders suggests lowering the transaction relay policy limit from 85 bytes to the 65 byte limit suggested in consensus cleanup, which may allow additional experimentation and usage without changing the current risk profile. Sanders has a pull request open to make this change. See also Newsletter #99 for prior discussion related to this proposed change.

  • BIP324 update: Dhruv M posted to the Bitcoin-Dev mailing list a summary of several updates to the BIP324 proposal for a version 2 encrypted P2P transport protocol. This includes a rewrite of the draft BIP and the publication of a variety of resources to help reviewers evaluate the proposal, including an excellent guide to the proposed code changes across multiple repositories.

    As described in the draft BIP’s motivation section, a native encrypted transport protocol for Bitcoin nodes can improve privacy during transaction announcement, prevent tampering with connections (or at least make it easier to detect tampering), and also make P2P connection censorship and eclipse attacks more difficult.

Changes to services and client software

In this monthly feature, we highlight interesting updates to Bitcoin wallets and services.

  • btcd v0.23.2 released: btcd v0.23.2 (and v0.23.1) adds addr v2 and additional support for PSBTs, taproot, and MuSig2 as well as other enhancements and fixes.

  • ZEBEDEE announces hosted channel libraries: In a recent blog post, ZEBEDEE announced an open source wallet (Open Bitcoin Wallet), Core Lightning plugin (Poncho), Lightning client (Cliché), and Lightning library (Immortan) which focus on support for hosted channels.

  • Cashu launches with Lightning support: E-cash software Cashu launches as a proof-of-concept wallet with Lightning receive support.

  • Address explorer Spiral launches: Spiral is an open source public address explorer that uses cryptography to provide privacy to users querying information about an address.

  • BitGo announces Lightning support: In a blog post, BitGo describes its custodial Lightning service that runs nodes on behalf of its clients and maintains payment channel liquidity.

  • ZeroSync project launches: The ZeroSync project is using Utreexo and STARK proofs to sync a Bitcoin node, as occurs in Initial Block Download (IBD).

Releases and release candidates

New releases and release candidates for popular Bitcoin infrastructure projects. Please consider upgrading to new releases or helping to test release candidates.

Notable code and documentation changes

Notable changes this week in Bitcoin Core, Core Lightning, Eclair, LDK, LND, libsecp256k1, Hardware Wallet Interface (HWI), Rust Bitcoin, BTCPay Server, BDK, Bitcoin Improvement Proposals (BIPs), and Lightning BOLTs.

  • Bitcoin Core #23549 adds the scanblocks RPC that identifies relevant blocks in a given range for a provided set of descriptors. The RPC is only available on nodes that maintain a compact block filter index (-blockfilterindex=1).

  • Bitcoin Core #25412 adds a new /deploymentinfo REST endpoint which contains information about soft fork deployments, similar to the existing getdeploymentinfo RPC.

  • LND #6956 allows configuring the minimum channel reserve enforced on payments received from a channel’s partner. A node won’t accept a payment from its channel partner if that would lower the amount of the partner’s funds in the channel below the reserve, which is 1% by default in LND. This ensures the partner will need to pay at least the reserve amount as a penalty if it attempts to close a channel in a outdated state. This merged PR allows lowering or raising the reserve amount.

  • LND #7004 updates the version of the BTCD library used by LND, fixing the security vulnerability previously described in this newsletter.

  • LDK #1625 begins tracking information about the liquidity of distant channels which the local node has attempted to route payments through. The local node stores information about the size of payments which have either successfully been routed through the remote node or which failed due to apparent insufficient funds. This information, adjusted for its age, is used as input for probabilistic pathfinding (see Newsletter #163).

Footnotes

  1. Transaction replacement was included in the first version of Bitcoin and has received much discussion over the years. During that time, several terms used for describing aspects of it have changed, leading to potential confusion. Perhaps the greatest source of confusion would be the term “full RBF”, which has been used for two different concepts:

    • Full replacement of any part of a transaction as distinct from just adding additional inputs and outputs. During a period when enabling RBF was controversial and before the idea of opt-in RBF was proposed, one suggestion was to allow a transaction to be replaced only if the replacement included all of the same outputs plus additional new inputs and outputs used to pay fees and collect change. The requirement to keep the original outputs ensured the replacement would still pay the original receiver the same amount of money. This idea, later called First Seen Safe (FSS) RBF, was a type of partial replacement.

      By comparison, full replacement at this time meant the replacement could fully change anything about the original transaction (provided it still conflicted with the original transaction by spending at least one of the same inputs). It’s this usage of full that’s used in the title of BIP125, “Opt-in Full Replace-by-Fee Signaling”.

    • Full replacement of any transaction as distinct from only replacing transactions that opt-in to allowing replacement via a BIP125 signal. Opt-in RBF was proposed as a compromise between people who didn’t want to allow RBF and those who believed it was either necessary or inevitable. However, as of this writing, only a minority of transactions opt-in to RBF, which can be seen as partial adoption of RBF.

      By comparison, full adoption of RBF can be enabled by allowing any unconfirmed transaction to be replaced. It’s this usage of full that’s used in the currently-discussed Bitcoin Core configuration option, mempoolfullrbf.