This week’s newsletter describes a new time warp that’s particularly consequential for the new testnet4, summarizes discussion about proposed mitigations for onion message denial-of-service concerns, seeks feedback on a proposal to allow LN payers to optionally identify themselves, and announces a major change to Bitcoin Core’s build system that could affect downstream developers and integrators. Also included are our regular sections announcing new releases and release candidates and describing notable changes to popular Bitcoin infrastructure software.

News

• ● New time warp vulnerability in testnet4: Mark “Murch” Erhardt posted to Delving Bitcoin to describe an attack discovered by developer Zawy for exploiting testnet4’s new difficulty adjustment algorithm. Testnet4 applied a solution from the consensus cleanup soft fork for mainnet that is intended to block the time warp attack. However, Zawy described an attack similar to time warp that could be used even with the new rule to reduce mining difficulty to 1/16th its normal value. Erhardt extended Zawy’s attack to allow reducing difficulty to its minimum value. We describe several related attacks in simplified form below:

  Bitcoin blocks are produced stochastically, with the difficulty intended to retarget every 2,016 blocks to keep the average time between blocks at about 10 minutes. The following simplified illustration shows what is supposed to happen with a constant rate of block production, given a retarget every 5 blocks (reduced from every 2,016 blocks to make the illustration legible):

  

  A dishonest miner (or cabal of miners) possessing slightly more than 50% of hashrate can censor blocks produced by the other slightly-less-than 50% of honest miners. That would naturally lead initially to only one block being produced every 20 minutes on average. After 2,016 blocks are produced using this pattern, difficulty will adjust to 1/2 its original value to return the rate of blocks being added to the mainchain to one block every 10 minutes on average:

  

  A time warp attack occurs when the dishonest miners use their hashrate majority to force the timestamps in most blocks to use the minimum allowed value. At the end of each 2,016-block retarget period, they increase block header time to the current wall time to make it seem like it took longer to produce the blocks than it actually did, leading to a lower difficulty in the subsequent period.

  

  The new rule applied to testnet4 fixes this by preventing the first block in a new retarget period from having a timestamp much earlier than its previous block (the last block in the previous period).

  Like the original time warp attack, Erhardt’s version of Zawy’s attack increments most block’s header time by the bare minimum. However, for two of every three retarget periods, it jumps forward the time for the final block in a period and the first block in the subsequent period. This decreases difficulty by the maximum allowed each period (1/4th the current value). For the third period, it uses the lowest time allowed for all blocks, plus the first block of the subsequent period, which increases difficulty by the maximum value (4x). In other words, difficulty decreases 1/4, decreases again to 1/16, and then increases to 1/4 of its original value:

  

  The three-period cycle can be repeated indefinitely to reduce difficulty by 1/4 each cycle, eventually reducing it to a level that allows miners to produce up to 6 blocks per second.

  To reduce difficulty by 1/4 in a retarget period, the attacking miners need to set the time of the retarget blocks to 8 weeks further in the future than the block at the start of the current period. To do this twice in a row at the start of the attack requires eventually setting some block’s time 16 weeks into the future. Full nodes will not accept blocks more than two hours into the future, preventing the malicious blocks from being accepted for 8 weeks for the first set of blocks and 16 weeks for the second set of blocks. While the attacking miners wait for their blocks to be accepted, they can create additional blocks at ever lower difficulties. Any blocks created by honest miners during the 16 weeks that attackers are preparing their attack will be reorganized out of the chain when full nodes begin accepting the attacker blocks; this could mark every transaction confirmed during that time as either unconfirmed or invalid (conflicted) on the current chain.

  Erhardt suggests solving the attack with a soft fork that requires the timestamp of the last block in a retarget period be greater than the timestamp of the first block in that period. Zawy proposed several solutions, including forbidding block timestamps from decreasing (which could create problems if some miners create blocks near the two-hour future limit enforced by nodes), or at least forbidding them from decreasing by more than about two hours.

  Overall, on mainnet, the new time warp attack is similar to the original attack in its requirements for mining equipment, its ability to be detected in advance, its consequences for users, and the relative simplicity of a soft fork to fix it. It depends on an attacker maintaining control over at least 50% of hashrate for at least a month, while likely signaling to users that an attack was impending and hoping they don’t respond, which could be quite challenging on mainnet. As Zawy notes, the attack is much easier on testnet: a small amount of modern mining equipment is enough to achieve 50% of hashrate on testnet and set up the attack in stealth. An attacker could then, in theory, produce over 500,000 blocks per day. Only someone willing to dedicate a greater amount of hashrate to testnet could stop an attacker unless the attack is prevented using a soft fork.

  At the time of writing, the tradeoffs between proposed solutions were being discussed.

• ● Onion message DoS risk discussion: Gijs van Dam posted to Delving Bitcoin to discuss a recent paper by researchers Amin Bashiri and Majid Khabbazian about onion messages. The researchers note each onion message can be forwarded across many nodes (481 hops by Van Dam’s calculations), potentially wasting bandwidth for all of those nodes. They describe several methods for reducing the risk of bandwidth abuse, including a clever method of requiring an exponentially increasing amount of PoW for each additional hop, making short routes computationally cheap but long routes expensive.

  Matt Corallo suggested first trying a previously proposed idea (see Newsletter #207) to provide back pressure on nodes sending too many onion messages before working on anything more complicated.

• ● Optional identification and authentication of LN payers: Bastien Teinturier posted to Delving Bitcoin to propose methods for allowing spenders to optionally include extra data with their payments that would allow receivers to identify those payments as having come from a known contact. For example, if Alice generates an offer that Bob pays, she may want cryptographic proof that the payment came from Bob and not from some third party pretending to be Bob. Offers are designed by default to hide the identities of the spender and receiver, so additional mechanisms are required to enable opt-in identification and authentication.

  Teinturier starts by describing an opt-in contact key distribution mechanism that Bob can use to disclose a public key of his to Alice. He then describes three potential candidates for an additional opt-in mechanism that Bob can use to sign his payments to Alice. If Bob uses that mechanism, Alice’s LN wallet can authenticate that signature as belonging to Bob and display that information to her. In unauthenticated payments, fields set by the spender (such as the free-form payer_note field) can be marked as untrusted to discourage users from relying on information provided in them.

  Feedback on which cryptographic methods to use is requested, with Teinturier planning to release BLIP42 with a specification for the methods selected.

• ● Bitcoin Core switch to CMake build system: Cory Fields posted to the Bitcoin-Dev mailing list to announce Bitcoin Core’s impending switch from the GNU autotools build system to the CMake build system, which has been led by Hennadii Stepanov with contributions from Michael Ford for bug fixes and modernization, with reviews and contributions from several other developers (including Fields). This should not affect anyone who uses the pre-built binaries available from BitcoinCore.org—which is what we expect most people are using. However, developers and integrators who build their own binaries for testing or customization may be affected—especially those working on uncommon platforms or build configurations.

  Fields’s email provides answers to anticipated questions and asks anyone who builds Bitcoin Core themselves to test PR #30454 and report any issues. That PR is expected to be merged in the next several weeks with anticipated release in version 29 (anticipated about 7 months from now). The earlier you test, the more time Bitcoin Core developers will have to fix problems before the release of version 29—increasing the chance that they can prevent problems in the released code from affecting your configuration.

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.

• ● BDK 1.0.0-beta.1 is a release candidate for this library for building wallets and other Bitcoin-enabled applications. The original bdk Rust crate has been renamed to bdk_wallet and lower layer modules have been extracted into their own crates, including bdk_chain, bdk_electrum, bdk_esplora, and bdk_bitcoind_rpc. The bdk_wallet crate “is the first version to offer a stable 1.0.0 API.”

• ● Core Lightning 24.08rc2 is a release candidate for the next major version of this popular LN node implementation.

• ● LND v0.18.3-beta.rc1 is a release candidate for a minor bug fix release of this popular LN node implementation.

Notable code and documentation changes

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

• ● Bitcoin Core #29519 resets the pindexLastCommonBlock value after an assumeUTXO snapshot is loaded, so that a node prioritizes downloading blocks after the most recent block in the snapshot. This fixes a bug where a node would set pindexLastCommonBlock from existing peers before loading the snapshot and start downloading blocks from that much-older block. Although older blocks still need to be downloaded for assumeUTXO’s background validation, newer blocks should receive priority so that users can see whether their recent transactions have been confirmed.

• ● Bitcoin Core #30598 removes block height from the assumeUTXO snapshot file metadata as it is not a unique identifier in a pre-sanitized untrusted file compared to the block hash that is kept. A node can still obtain block height from many other internal sources.

• ● Bitcoin Core #28280 optimizes Initial Block Download (IBD) performance for pruned nodes by not emptying the UTXO cache during pruning flushes. It does this by separately tracking “dirty” cache entries—entries that have changed since they were last written to the database. This allows a node to avoid unnecessarily scanning the entire cache during prune flushes and instead focus on the dirty entries. The optimization results in up to 32% faster IBD for pruned nodes with high dbcache settings, and about 9% improvement with default settings. See Newsletter #304.

• ● Bitcoin Core #28052 adds XOR encoding to blocksdir *.dat files on creation as a preventative mechanism against unintentional and accidental data corruption by anti-virus or similar software. This can be optionally disabled and doesn’t protect against intentional data corruption attacks. This was implemented for the chainstate files in Bitcoin Core #6650 in September 2015 and the mempool in #28207 in November 2023 (see Newsletter #277).

• ● Core Lightning #7528 adjusts the fee rate estimation for sweeps of non-time-sensitive unilateral channel closures to an absolute deadline of 2016 blocks (approximately 2 weeks). Previously, the fee rate was set to target confirmation within 300 blocks, which sometimes caused transactions to be stuck at the minimum relay fee limit, resulting in indefinite delays.

• ● Core Lightning #7533 updates the internal coin movement notifications and the transaction bookkeeper to properly account for the spending of funding outputs for splicing transactions. Previously, there was no logging or tracking of this.

• ● Core Lightning #7517 introduces askrene, a new experimental plugin and API infrastructure for minimum-cost pathfinding based on the renepay plugin (See Newsletter #263) for an improved implementation of Pickhart Payments. The getroutes RPC command allows the specification of persistent channel capacity data and transient information such as blinded paths or route hints, and returns a set of possible routes along with their estimated probability of success. Several other RPC commands are added to manage routing data in layers by adding channels, manipulating channel data, excluding nodes from routing, inspecting layer data, and managing ongoing payment attempts.

• ● LND #8955 adds an optional utxo field on the sendcoins command (and Outpoints for the corresponding SendCoinsRequest RPC command) to simplify the coin selection user experience to a single step. Previously, a user had to go through a multi-step command process that included coin selection, fee estimation, PSBT funding, PSBT completion, and transaction broadcasting.

• ● LND #8886 updates the BuildRoute function to support inbound forwarding fees by reversing the pathfinding process, now working from receiver to sender, allowing for more accurate fee calculations over multiple hops. See Newsletter #297 for more on inbound fees.

• ● LND #8967 adds a new wire message type Stfu (SomeThing Fundamental is Underway) designed to lock the channel state before initiating protocol upgrades. The Stfu message type includes fields for the channel id, an initiator flag, and additional data for potential future extensions. This is part of the Quiescence protocol implementation (see Newsletter #309).

• ● LDK #3215 checks that a transaction is at least 65 bytes long to protect against an unlikely and costly attack against a light client SPV wallet where a valid SPV proof can be created for a fake 64-byte transaction by matching the hash of an inner merkle node. See merkle tree vulnerabilities.

• ● BLIPs #27 adds BLIP04 for an experimental HTLC endorsement signaling protocol proposal to partially mitigate channel jamming attacks on the network. It outlines the experimental endorsement TLV values, the deployment approach, and the eventual deprecation of the experimental phase when HTLC endorsements are merged into the BOLTs.