mirror of
https://github.com/bitcoin/bips.git
synced 2026-03-02 15:43:53 +00:00
* new bip: timelock recovery storage format * Comparison with Script-Based Wallets * Type is Specification Co-authored-by: Mark "Murch" Erhardt <murch@murch.one> * Change Authors to a single Author Co-authored-by: Mark "Murch" Erhardt <murch@murch.one> * Replace OP_VAULT mention with OP_CHECKCONTRACTVERIFY * Only the Alert Transaction needs to be non-malleable * Adding discussion link * limiting the transactions weight This is important in order to prevent users from creating recovery-plans that are hard to propagate. * Explain anchor-addresses * fix typo Co-authored-by: Mark "Murch" Erhardt <murch@murch.one> * add surname initial to author name * Explain unintentional initiation of rrecovery-plan. * limit alert_inputs length to 2439 * updating bip number to 128 * rename to bip-0128.mediawiki * BIP 128: Timelock-Recovery storage format * fix field order, change title to uppercase * Making plugin_version optional Relevant only in wallets where the feature is implemented via a plugin. * Removing mainnet Irrelevant. Obviously a monitoring service for mainnet should verify that the addresses are on mainnet.
267 lines
18 KiB
Plaintext
267 lines
18 KiB
Plaintext
<pre>
|
|
BIP: 128
|
|
Layer: Applications
|
|
Title: Timelock-Recovery Storage Format
|
|
Authors: Oren Z <orenz0@protonmail.com>
|
|
Status: Draft
|
|
Type: Specification
|
|
Assigned: 2026-02-05
|
|
License: BSD-2-Clause
|
|
Discussion: https://groups.google.com/g/bitcoindev/c/K1NpJp9_BYk
|
|
</pre>
|
|
|
|
== Abstract ==
|
|
|
|
This document proposes a standard format for saving timelock-recovery plans, to allow different
|
|
wallets to generate them, and different services to monitor/execute them.
|
|
|
|
== Motivation ==
|
|
|
|
Pre-signed transactions are one way to create a recovery-plan, for use in case of seed loss or
|
|
inheritance.
|
|
The most common example is a single pre-signed transaction with an <code>nLocktime</code> set to a
|
|
future date, as explained in [[bip-0065.mediawiki|BIP-65]].
|
|
One limitation of this approach is that in the happy-flow scenario, when the seed is not lost,
|
|
and the <code>nLocktime</code> is about to be reached, the user must access their wallet and spend
|
|
one of its UTXOs - in order to revoke the pre-signed transaction and prevent it from being able to
|
|
move the funds with no cancellation period.
|
|
This could be frustrating, for example, for users that split their seed over multiple geographic
|
|
locations.
|
|
|
|
''Timelock-Recovery plans'' are a way to pre-sign a pair of transactions that eventually move the
|
|
funds to one or more secondary wallets - with a special <code>nSequence</code> relative-locktime
|
|
in the second transaction, so that the user always has a cancellation-period.
|
|
|
|
Executing and monitoring a ''Timelock-Recovery plan'' thus requires more than broadcasting and
|
|
monitoring a single transaction. It also requires mechanisms for accelerating the first
|
|
transaction (which does not move most funds to the secondary wallet), for checking whether
|
|
the relative-timelock has passed, and a more nuanced handling of reorgs.
|
|
|
|
This BIP proposes a standard format for exporting ''Timelock-Recovery plans'' from the wallet that
|
|
generated them, and importing them into apps/services for monitoring/execution.
|
|
|
|
=== Comparison with Script-Based Wallets ===
|
|
|
|
Script-based wallets are another way to create recovery mechanisms, and can use absolute and
|
|
relative locktimes using OP_CHECKLOCKTIMEVERIFY ([[bip-0065.mediawiki|BIP-65]]) and
|
|
OP_CHECKSEQUENCEVERIFY ([[bip-0112.mediawiki|BIP-112]]).
|
|
For example, we can build a script that allows one main key to spend the funds at any time,
|
|
and a secondary key to spend the funds only in transactions with nLocktime above a certain
|
|
date/block-height, or only in transactions with nSequence above a certain relative
|
|
time-gap/number-of-blocks.
|
|
This makes the secondary key useful only after an absolute date/block-height, or after
|
|
a relative time since the funds were received (each UTXO independently).
|
|
This approach does have some advantages over pre-signed transactions, for example the
|
|
recovery-mechanism automatically applies to new funds received into the wallet.
|
|
|
|
However, script-based wallets have some disadvantages over a sequence of
|
|
pre-signed transactions:
|
|
|
|
* Script-based wallets are harder to implement correctly by hardware wallets, and harder to backup properly (i.e. users may forget to backup wallet-descriptors even for basic multisig wallets).
|
|
* As of the time of writing, scripts can limit when secondary-keys can be used, but not how they can be used: if the user doesn't touch the wallets' UTXOs for long-enough time, the secondary key will eventually become useable and could move the funds anywhere. This is true whether we measure the time in absolute terms (OP_CHECKLOCKTIMEVERIFY) or relative terms compared to when the wallets' UTXOs were created (OP_CHECKSEQUENCEVERIFY). This means that even in the happy-flow scenario of an untouched wallet, where no recovery is needed, the user must periodically "renew" the recovery-mechanism by spending the UTXO to a new wallet/address. This may be inconvenient in ultra-cold-storage scenarios (i.e. multisig with main keys hidden in different geographic locations). New opcode suggestions, such as OP_CHECKTEMPLATEVERIFY ([[bip-0119.mediawiki|BIP-119]]) and OP_CHECKCONTRACTVERIFY ([[bip-0443.mediawiki|BIP-443]]), discuss possible recovery-mechanisms in which in order for a secondary key to have full control over the funds, some onchain operations must be performed, with a required time-gap between them - giving the user enough time to revoke the whole process and move the funds elsewhere (assuming they still have the main key and the recovery-mechanism was triggered unintentionally). However, these suggestions are still in the discussion phase and even if ever implemented, their adoption may be slow.
|
|
* New Bitcoiners today typically don't think of such recovery-mechanisms in advance, and start with a P2WPKH wallet. They can pre-sign transactions with this wallet, but to utilize script-based features they would need to create a new wallet and move the funds there - an operation that might seem intimidating for large amounts.
|
|
|
|
== Specification ==
|
|
|
|
A ''Timelock-Recovery plan'' consists of two transactions:
|
|
|
|
* ''Alert Transaction'': A mostly-consolidation transaction that keeps most funds in the original wallet, except for a fee and a small fixed amount that goes to ''anchor-addresses'' - addresses which can be used to accelerate the ''Alert Transaction'' via CPFP. The majority of funds should remain on the original wallet, in a new previously-unused address which we call the ''alert-address''. We use the term ''Alert Transaction'' because monitoring the blockchain and looking for it should alert the user that the recovery-plan has been initiated (intentionally, unintentionally or maliciously).
|
|
* ''Recovery Transaction'': The transaction that moves the funds from the alert-address UTXO from the ''Alert Transaction'' to one or more addresses of secondary wallets (each may receive a different amount). This transaction should have a special <code>nSequence</code> relative-locktime according to the size of cancellation-period requested by the user, following the rules of [[bip-0068.mediawiki|BIP-68]].
|
|
|
|
With a reliable tool to monitor the blockchain for the ''Alert Transaction''
|
|
or the ''Alert Address'', the user can safely store online backups of the recovery plan's
|
|
JSON file (or, even without a tool, by checking the blockchain manually from time to time).
|
|
If the presigned transactions leak and the ''Alert Transaction'' is broadcast
|
|
unintentionally, the user has the cancellation period (expected to be at least a
|
|
few days) to prevent most funds from moving by sending them to a new address, thereby
|
|
invalidating the ''Recovery Transaction''.
|
|
|
|
It is important that the ''Alert Transaction'' will be non-malleable (e.g. by using
|
|
[[bip-0140.mediawiki|BIP-140]]).
|
|
If a malleable ''Alert Transaction'' is used, a malicious miner could replace the
|
|
''Alert Transaction'' with a similar transaction with a different txid,
|
|
making the ''Recovery Transaction'' invalid (pointing to a non-existent UTXO).
|
|
|
|
The <code>nLocktime</code> of both transactions should not be higher than the current
|
|
block height.
|
|
|
|
The ''anchor-addresses'' mentioned above, which are used for CPFP acceleration, could possibly
|
|
be P2A addresses (described in [[bip-0433.mediawiki|BIP-433]]), or other addresses under the
|
|
participants' control (i.e. addresses from the secondary wallets).
|
|
As of the time of writing, P2A is not widely adopted, and less-technical users may
|
|
struggle using them for CPFP acceleration - so we currently recommend using regular addresses.
|
|
|
|
=== nSequence calculation ===
|
|
|
|
Users will specify the cancellation-period in whole days between 2-388.
|
|
|
|
Following [[bip-0068.mediawiki|BIP-68]], the <code>nSequence</code> can represent a timespan in
|
|
units of 512 seconds, when bit (1 << 22) is set. An example calculation is provided below:
|
|
|
|
<source lang="python">
|
|
n_sequence = (1 << 22) | round(cancellation_period_days * 24 * 60 * 60 / 512)
|
|
</source>
|
|
|
|
Users should be notified that the cancellation-period is not guaranteed to be exact (due to miners'
|
|
manipulation of block-timestamps).
|
|
|
|
Less than 2 days of cancellation-period and partial-days are not supported, as they are not useful.
|
|
|
|
More than 388 days of cancellation-period will overflow the <code>nSequence</code> field bits
|
|
allocated for the relative-locktime, and is not supported.
|
|
|
|
=== JSON format ===
|
|
|
|
For simplicity, this BIP proposes that a ''Timelock-Recovery plan'' will be saved as a JSON
|
|
object.
|
|
|
|
The JSON object will have the following fields:
|
|
|
|
* kind (mandatory): must be "timelock-recovery-plan".
|
|
* id (mandatory): a non-empty string of up to 100 characters, to represent the plan uniquely (i.e. a UUID, or a server generated ID).
|
|
* name (optional): a name for the plan, decided by the user. A string of up to 200 characters.
|
|
* description (optional): a description for the plan, decided by the user. A string of up to 10,000 characters.
|
|
* created_at (mandatory): an ISO 8601 timestamp of the plan creation time, including timezone offset ('Z' if the timezone is UTC).
|
|
* plugin_version (optional): The version of the plugin that generated the plan. A string of up to 100 characters.
|
|
* wallet_version (mandatory): The version of the wallet that generated the plan. A string of up to 100 characters.
|
|
* wallet_name (mandatory): The human-readable name of the wallet app that generated the plan. A string of up to 100 characters.
|
|
* wallet_kind (mandatory): The internal name of the wallet app that generated the plan. A string of up to 100 characters.
|
|
* timelock_days (mandatory): The cancellation period in whole days. A number between 2 and 388.
|
|
* anchor_amount_sats (mandatory): The amount in satoshis sent to each anchor address in the <code>Alert Transaction</code>. We recommend using 600 sats, which is above the dust limit.
|
|
* anchor_addresses (mandatory): An array of up to 10,000 Bitcoin addresses that receive the anchor amount in the <code>Alert Transaction</code>. Each address is a string of up to 100 characters.
|
|
* alert_address (mandatory): The Bitcoin address that receives the majority of funds in the <code>Alert Transaction</code>. A string of up to 100 characters.
|
|
* alert_inputs (mandatory): An array of up to 2439 inputs spent by the <code>Alert Transaction</code>. Each input is a string in the format "txid:vout" where txid is a 64-character lowercase hexadecimal string and vout is a decimal number of up to 6 digits. The maximal length of 2439 is calculated from a standard transaction of 400,000 wu where each input contains at least 41 bytes.
|
|
* alert_tx (mandatory): The raw <code>Alert Transaction</code> in uppercase hexadecimal format. A string of up to 800,000 characters.
|
|
* alert_txid (mandatory): The transaction ID of the <code>Alert Transaction</code>. A 64-character lowercase hexadecimal string.
|
|
* alert_fee (mandatory): The total fee paid by the <code>Alert Transaction</code> in satoshis. A non-negative integer.
|
|
* alert_weight (mandatory): The weight of the <code>Alert Transaction</code>. A positive integer, not higher than 400,000.
|
|
* recovery_tx (mandatory): The raw <code>Recovery Transaction</code> in uppercase hexadecimal format. A string of up to 800,000 characters.
|
|
* recovery_txid (mandatory): The transaction ID of the <code>Recovery Transaction</code>. A 64-character lowercase hexadecimal string.
|
|
* recovery_fee (mandatory): The total fee paid by the <code>Recovery Transaction</code> in satoshis. A non-negative integer.
|
|
* recovery_weight (mandatory): The weight of the <code>Recovery Transaction</code>. A positive integer, not higher than 400,000.
|
|
* recovery_outputs (mandatory): An array of up to 10,000 outputs from the <code>Recovery Transaction</code>. Each output is a tuple containing: <code>[address, amount_sats, label?]</code> where:
|
|
** address is a mandatory Bitcoin address string (up to 100 characters).
|
|
** amount_sats is a mandatory positive integer representing the amount in satoshis.
|
|
** label is an optional string of up to 200 characters.
|
|
* metadata (optional): A string of up to 10,000 characters for additional metadata, for example a digital-signature.
|
|
* checksum (mandatory): A checksum for verifying the integrity of the plan. A string of 8 to 64 characters.
|
|
|
|
=== Checksum Calculation ===
|
|
Notice that besides the top-level JSON object, all the internal values are either primitive or
|
|
arrays.
|
|
This is intentional, so a conversion of the values to JSON strings will be deterministic.
|
|
|
|
The checksum is calculated by converting the top-level JSON object to an array of
|
|
<code>[key, value]</code> pairs, sorting the array, stringifying, calculating the
|
|
SHA256 hash of the result in lowercase hexadecimal format, and taking a prefix of at least 8
|
|
characters.
|
|
|
|
For example:
|
|
<source lang="javascript">
|
|
const checksumData = new TextEncoder().encode(
|
|
JSON.stringify(Object.entries(recoveryPlanJson).sort()),
|
|
);
|
|
const checksum = new Uint8Array(await crypto.subtle.digest('SHA-256', checksumData));
|
|
const checksumHex = Array.from(checksum).map(b => b.toString(16).padStart(2, '0')).join('').slice(0, 8);
|
|
</source>
|
|
|
|
Checksum hex string should be at least 8 characters long. Wallets may choose to use a longer
|
|
checksum.
|
|
|
|
== Rationale ==
|
|
|
|
The JSON object will contain the raw transactions, in addition to other information - some of
|
|
which could technically be extracted from the raw transactions. This is intentional, to let
|
|
frontend UIs display the plan before uploading it to any service, without the need for
|
|
complicated parsing in the frontend.
|
|
|
|
Backend services that receive the JSON object for monitoring/execution are expected to validate
|
|
that the information is consistent with the raw transactions.
|
|
|
|
Also, if some wallet apps did not implement the specifications correctly, the services could
|
|
write custom code based on the <code>wallet_kind</code>, <code>wallet_version</code> and
|
|
<code>plugin_version</code> fields.
|
|
|
|
Servers may decide to put more restrictions on JSON objects, for example to refuse
|
|
storing very large transactions.
|
|
|
|
Notice that the raw transactions (<code>alert_tx</code> and <code>recovery_tx</code>) are expected
|
|
to be in uppercase hexadecimal format.
|
|
This is useful for frontend UIs to display them as QR codes, which are more compact when using
|
|
uppercase-only alphanumeric characters.
|
|
|
|
=== Monitoring Timelock-Recovery Plans ===
|
|
|
|
Checking whether the <code>Alert Transaction</code> is valid is trivial, via the
|
|
<code>testmempoolaccept</code> RPC call in bitcoin core 0.17+.
|
|
|
|
However, checking whether the <code>Recovery Transaction</code> is valid is more complex,
|
|
since it depends on a UTXO created by the <code>Alert Transaction</code>.
|
|
|
|
The <code>testmempoolaccept</code> RPC can receive a list of transactions in which the later
|
|
transactions may depend on earlier transactions - however in our case the
|
|
<code>Recovery Transaction</code> has an <code>nSequence</code> relative-locktime, and therefore
|
|
calling <code>testmempoolaccept 'alert-tx' 'recovery-tx'</code> will fail, claiming that the
|
|
<code>Alert Transaction</code> UTXO is not confirmed (and the required time window has not passed).
|
|
|
|
We recommend services that want to verify the entire <code>Timelock-Recovery plan</code> to parse
|
|
the <code>Recovery Transaction</code> and check its signatures manually, and reject complicated
|
|
spending scripts. Discovering that the <code>Recovery Transaction</code> is invalid only at the
|
|
time of execution, could lead to funds being locked forever.
|
|
|
|
== Reference Implementation ==
|
|
|
|
JSON files can be generated using the Timelock Recovery plugin on
|
|
[https://electrum.org Electrum Wallet]:
|
|
|
|
https://github.com/spesmilo/electrum/tree/master/electrum/plugins/timelock_recovery
|
|
|
|
Demo Video: https://drive.google.com/file/d/10uXRouQbH1kz_HC14WnmRnYHa3gPZY8l/preview
|
|
|
|
Example JSON file:
|
|
|
|
<source lang="json">
|
|
{
|
|
"kind": "timelock-recovery-plan",
|
|
"id": "exported-692452189b301b561ed57cbe",
|
|
"name": "Recovery Plan ac300e72-7612-497e-96b0-df2fdeda59ea",
|
|
"description": "RITREK APP 1.1.0: Trezor Account #1",
|
|
"created_at": "2025-11-24T12:39:53.532Z",
|
|
"plugin_version": "1.0.1",
|
|
"wallet_version": "1.0.1",
|
|
"wallet_name": "RITREK Service",
|
|
"wallet_kind": "RITREK BACKEND",
|
|
"timelock_days": 2,
|
|
"anchor_amount_sats": 600,
|
|
"anchor_addresses": [
|
|
"bc1qnda6x2gxdh3yujd2zjpsd7qzx3awxmlaf9wwlk"
|
|
],
|
|
"alert_address": "bc1qj0f9sjenwyjs0u7mlgvptjp05z3syzq7mru3ep",
|
|
"alert_inputs": [
|
|
"a265a485df4c6417019b91379257eb387bceeda96f7bb6311794b8ed358cf104:0",
|
|
"2f621c2151f33173983133cbc1000e3b603b8a18423b0379feffe8513171d5d3:0"
|
|
],
|
|
"alert_tx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
|
|
"alert_txid": "f1413fedadaf30697820bcd8f6a393fcc73ea00a15bea3253f89d5658690d2f7",
|
|
"alert_fee": 231,
|
|
"alert_weight": 834,
|
|
"recovery_tx": "02000000000101F7D2908665D5893F25A3BE150AA03EC7FC93A3F6D8BC20786930AFADED3F41F101000000005201400001A6550000000000001600149B7BA329066DE24E49AA148306F802347AE36FFD0247304402204AFF87C2127F5697F300C6522067A8D5E5290CA8D140D2E5BCEF4A36606C5FE5022056673BEC5BB459DFFBD4D266EE95AEF0D701383ED80BD433A02C3C486A826D76012102774DBCD59F2D08EFF718BC09972ADC609FBC31C26B551B3E4EA30A1D43EEDB9700000000",
|
|
"recovery_txid": "bc304610e8f282036345e87163d4cba5b16488a3bf2e4d738379d7bda3a0bca3",
|
|
"recovery_fee": 122,
|
|
"recovery_weight": 437,
|
|
"recovery_outputs": [
|
|
[
|
|
"bc1qnda6x2gxdh3yujd2zjpsd7qzx3awxmlaf9wwlk",
|
|
21926,
|
|
"My Backup Wallet"
|
|
]
|
|
],
|
|
"metadata": "sig:825d6b3858c175c7fc16da3134030e095c4f9089c3c89722247eeedc08a7ef4f",
|
|
"checksum": "92f8b3da"
|
|
}
|
|
</source>
|
|
|
|
== Copyright ==
|
|
|
|
This document is licensed under the 2-clause BSD license.
|