public inbox for bitcoindev@googlegroups.com
 help / color / mirror / Atom feed
From: Nicolas Dorier <nicolas.dorier@gmail.com>
To: Bitcoin Dev <bitcoin-dev@lists.linuxfoundation.org>
Subject: [bitcoin-dev] BIP Number request for a simple payjoin proposal
Date: Sun, 17 May 2020 04:46:35 +0900	[thread overview]
Message-ID: <CA+1nnrkq2zq_PFoUyNadE7KMncbv30M=Kr-Ek4nDfnJvgDHrfQ@mail.gmail.com> (raw)

[-- Attachment #1: Type: text/plain, Size: 21041 bytes --]

I am requesting a BIP number to be allocated for this simple payjoin proposal.
This proposal is already being implemented by several service and
wallets and incorporate the feedback of the community at
https://github.com/NicolasDorier/bips/pull/3

I opened a pull request at: https://github.com/bitcoin/bips/pull/923

I am not checking my mail very often, so I suggest give me feedback
directly on the opened pull request.


<pre>
  BIP: ?
  Layer: Applications
  Title: A Simple Payjoin Proposal
  Author: Nicolas Dorier <nicolas.dorier@gmail.com>
  Comments-Summary: No comments yet.
  Comments-URI: https://github.com/bitcoin/bips/wiki/Comments:BIP-X
  Status: Draft
  Type: Standards Track
  Created: 2019-05-01
  License: BSD-2-Clause
</pre>

==Introduction==

===Abstract===

This document proposes a protocol for two parties
to negotiate a coinjoin transaction during a payment between them.

===Copyright===

This BIP is licensed under the 2-clause BSD license.

===Motivation===

When two parties (later referred to as sender and receiver) want to transact,
most of the time, the sender creates a transaction spending their own
Unspent Transaction Outputs (UTXOs), signs
it and broadcasts it on the network.

This simple model gave birth to several heuristics impacting the
privacy of the parties and of the network as a whole.

* Common input ownership heuristic: In most transactions, all the
inputs belong to the same party.
* Change identification from scriptPubKey type: If all inputs are
spending UTXOs of a certain scriptPubKey type, then the change output
is likely to have the same scriptPubKey type, too.
* Change identification from round amount: If an output in the
transaction has a round amount, it is likely an output belonging to
the receiver.

We will designate these three heuristics as <code>common-input</code>,
<code>change-scriptpubkey</code>, <code>change-round-amount</code>.

The problems we aim to solve are:
* For the receiver, there is a missed opportunity to consolidate their
own UTXOs or making payment in the sender's transaction.
* For the sender, there are privacy leaks regarding their wallet that
happen when someone applies the heuristics detailed above to their
transaction.

Our proposal gives an opportunity for the receiver to consolidate
their UTXOs while also batching their own payments, without creating a
new transaction. (Saving fees in the process)
For the sender, it allows them to invalidate the three heuristics
above. With the receiver's involvement, the heuristics can even be
poisoned. (ie, using the heuristics to intentionally mislead
blockchain analysis)

Note that the existence of this proposal is also improving the privacy
of parties who are not using it by making the three heuristics
unreliable to the network as a whole.

=== Relation to BIP79 (Bustapay) ===

Another implementation proposal has been written:
[[https://github.com/bitcoin/bips/blob/master/bip-0079.mediawiki|BIP79
Bustapay]].

We decided to deviate from it for several reasons:
* It was not using PSBT, so if the receiver wanted to bump the fee,
they would need the full UTXO set.
* The receiver was responsible to pay the additional fee, not the sender.
* It was requiring at least one input to be contributed by the receiver.
* Inability to change the payment output to match scriptPubKey type.
* Lack of basic versioning negotiation if the protocol evolves.
* No standardization of error condition for proper feedback to the sender.

Other than that, our proposal is very similar.

==Specification==

===Protocol===

In a payjoin payment, the following steps happen:

* The receiver of the payment, presents a [[bip-021.mediawiki|BIP 21
URI]] to the sender with a parameter <code>pj</code> describing an
https (or http if it is a Tor hidden service) link to the payjoin
endpoint.
* The sender creates a signed, finalized PSBT with witness UTXO or
previous transactions of the inputs. We call this PSBT the
<code>original</code>.
* The receiver replies back with a signed PSBT containing his own
signed inputs/outputs and those of the sender. We call this PSBT
<code>Payjoin proposal</code>.
* The sender verifies the proposal, re-signs his inputs and broadcasts
the transaction to the Bitcoin network. We call this transaction
<code>Payjoin transaction</code>.
<pre>
+----------+                        +--------+         +-----------------+
| Receiver |                        | Sender |         | Bitcoin Network |
+----+-----+                        +---+----+         +-------+---------+
     |       +-----------------+        |                      |
     +-------+ BIP21 with ?pj= +------->+                      |
     |       +-----------------+        |                      |
     |                                  |                      |
     |        +---------------+         |                      |
     +<-------+ Original PSBT +---------+                      |
     |        +---------------+         |                      |
     |                                  |                      |
     |       +------------------+       |                      |
     |       | Payjoin Proposal |       |                      |
     +-------+      PSBT        +------>+                      |
     |       +------------------+       |                      |
     |                                  |   +--------------+   |
     |                                  |---+ Payjoin      |   |
     |                                  |   | transaction  +-->+
     |                                  |   +--------------+   |
     +                                  +                      +
</pre>
The original PSBT is sent in the HTTP POST request body, base64
serialized, with <code>text/plain</code> in the
<code>Content-Type</code> HTTP header and <code>Content-Length</code>
set correctly.
The payjoin proposal PSBT is sent in the HTTP response body, base64
serialized with HTTP code 200.

To ensure compatibility with web-wallets and browser-based-tools, all
responses (including errors) must contain the HTTP header
<code>Access-Control-Allow-Origin: *</code>.

The sender must ensure that the url refers to a scheme or protocol
using authenticated encryption, for example TLS with certificate
validation, or a .onion link to a hidden service whose public key
identifier has already been communicated via a TLS connection. Senders
MUST NOT accept a url representing an unencrypted or unauthenticated
connection.

===Receiver's well known errors===

If for some reason the receiver is unable to create a payjoin
proposal, it will reply with a HTTP code different than 200.
The receiver is not constrained to specific set of errors, some are
specified in this proposal.

The errors have the following format:
<pre>
{
    "errorCode": "leaking-data",
    "message": "Key path information or GlobalXPubs should not be
included in the original PSBT."
}
</pre>

The well-known error codes are:
{| class="wikitable"
!Error code
!Meaning
|-
|leaking-data
|Key path information or GlobalXPubs should not be included in the
original PSBT.
|-
|psbt-not-finalized
|The original PSBT must be finalized.
|-
|unavailable
|The payjoin endpoint is not available for now.
|-
|out-of-utxos
|The receiver does not have any UTXO to contribute in a payjoin proposal.
|-
|not-enough-money
|The receiver added some inputs but could not bump the fee of the
payjoin proposal.
|-
|insane-psbt
|Some consistency check on the PSBT failed.
|-
|version-unsupported
|This version of payjoin is not supported.
|-
|need-utxo-information
|The witness UTXO or non witness UTXO is missing
|-
|invalid-transaction
|The original transaction is invalid for payjoin
|}

The receiver is allowed to return implementation specific errors which
may assist the sender to diagnose any issue.

However, it is important that error codes that are not well-known and
that the message do not appear on the sender's software user
interface.
Such error codes or messages could be used maliciously to phish a non
technical user.
Instead those errors or messages can only appear in debug logs.

It is advised to hard code the description of the error codes into the
sender's software.

===Receiver's original PSBT checklist===

The receiver needs to do some check on the original PSBT before proceeding:

* Non-interactive receivers (like a payment processor) need to check
that the original PSBT is broadcastable. <code>*</code>
* If the sender included inputs in the original PSBT owned by the
receiver, the receiver must either return error
<code>invalid-transaction</code> or make sure they do not sign those
inputs in the payjoin proposal.
* If the sender's inputs are all from the same scriptPubKey type, the
receiver must match the same type. If the receiver can't match the
type, they must return error <code>out-of-utxos</code>.

<code>*</code>: Interactive receivers are not required to validate the
original PSBT because they are not exposed to probing attacks.

===Sender's payjoin proposal checklist===

The sender should check the payjoin proposal before signing it to
prevent a malicious receiver from stealing money.

* Check that all the spent outpoints in the original PSBT still exist
in the coinjoin PSBT.
* Check that all the spent outpoints in the original PSBT do not have
any partial signature.
* If the sender is not using inputs with mixed types, check that the
receiver inputs type match the inputs type of the sender. (ie. both
using P2SH-P2WPKH or both using P2WPKH)
* Check that any inputs added by the receiver are finalized.
* Check that the transaction version, and nLockTime are unchanged.
* Check that the sender's inputs' sequence numbers are unchanged.
* If the sender's inputs' sequence numbers the homogenous, check that
the receiver's contributed inputs match those.
* Check that the sender's outputs have not been modified (but
potentially shuffled), except for paying increased fee
* If sender specified <code>feebumpindex=</code> (see later), the fee
should have been subtracted from the output at the same index in the
original PSBT.
* Check that the sent amount in the payjoin proposal is less than or
equal to the sent amount of the original transaction.

If the sent amount in the payjoin proposal is above the amount sent in
the original PSBT
* Check that the additional paid amount has been add paid to the fee.
* Check that the estimated fee rate of the payjoin proposal is not
more than the fee rate of the original PSBT. (fee estimation is hard,
so we should allow ~2 satoshi per inputs as margin of error)
* If <code>maxfeebumpcontribution=</code> was specified, check the
additional paid amount is less than or equal to this amount.
* If <code>maxfeebumpcontribution=</code> was not specified, the
sender's software should ask an interactive confirmation to the user.

The sender must be careful to only sign the inputs that were present
in the original PSBT and nothing else.

Note:
* The sender should allow the payment output to be modified by the
receiver (The receiver may substitute a P2WPKH payment to P2SH payment
to increase privacy)
* The sender must allow the receiver to add outputs.
* The sender must allow the receiver to not add any input. Useful for
the receiver to change the paymout output scriptPubKey type.
* If no input has been added, the sender's wallet should accept the
payjoin proposal, but should not mark the transaction as an actual
payjoin in the user interface.

Our method of checking the fee allows the receiver and the sender to
batch payments in the payjoin transaction.
It also allows the receiver to pay the fee for batching adding his own outputs.

===Optional parameters===

When the payjoin sender posts the original PSBT to the receiver, he
can optionally specify the following HTTP query string parameters:

* <code>v=</code>, the version number of the payjoin protocol that the
sender is using. The current version is <code>1</code>.

This can be used in the future so the receiver can reject a payjoin if
the sender is using a version which is not supported via an error HTTP
400, <code>version-unsupported</code>.
If not specified, the receiver will assume the sender is <code>v=1</code>.

If the receiver does not support the version of the sender, they
should send an error with the list of supported versions:
<pre>
{
    "errorCode": "version-unsupported",
    "supported" : [ 2, 3, 4 ],
    "message": "The version is not supported anymore"
}
</pre>

* <code>feebumpindex=</code>, the preferred output from which to
increase the fee for the added inputs. (default: <code>-1</code>)

If the <code>feebumpindex</code> is out of bounds or pointing to the
payment ouptut meant for the receiver, the receiver should ignore the
parameter.

* <code>maxfeebumpcontribution=</code>, an integer defining the
maximum amount in satoshis that the sender is willing to contribute
towards fees for the additional inputs.
<code>maxfeebumpcontribution</code> must be ignored if set to less
than zero. (default: -1)

Note that if <code>maxfeebumpcontribution</code> is too low, the
sender should create a transaction with RBF disabled, as the original
transaction could replace the payjoin transaction.

==Rationale==

There is several consequences of our proposal:

* The receiver can bump the fee of the original transaction.
* The receiver can modify the outputs of the original PSBT.
* The sender must provide the UTXO information (Witness or previous
transaction) in the PSBT.

===Respecting the minimum relay fee policy===

To be properly relayed, a Bitcoin transaction needs to pay at least 1
satoshi per virtual byte.
When fees are low, the original transaction is already 1 satoshi per
virtual byte, so if the receiver adds their own input, they need to
make sure the fee is increased such that the rate does not drop below
1 satoshi per virtual byte.

===Preventing mempool replacement===

A safe way to implement payjoin, is for both the sender and receiver
to try broadcasting the original transaction at some fixed interval
period regardless of the state of the payjoin.

If the receiver was not properly adding fees to the payjoin
transaction, the original transaction would end up replacing the
payjoin transaction in the mempool.

===Defeating heuristics based on the fee calculation===

Most wallets are creating a round fee rate (like 2 sat/b).
If the payjoin transaction's fee was not increased by the added size,
then those payjoin transactions could easily be identifiable on the
blockchain.

Not only would those transactions stand out by not having a round fee
(like 1.87 sat/b), but any suspicion of payjoin could be confirmed by
checking if removing one input would create a round fee rate.

===Receiver does not need to be a full node===

Because the receiver needs to bump the fee to keep the same fee rate
as the original PSBT, it needs the input's UTXO information to know
what is the original fee rate. Without PSBT, light wallets like Wasabi
Wallet would not be able to receive a payjoin transaction.

The validation (policy and consensus) of the original transaction is
optional: a receiver without a full node can decide to create the
payjoin transaction and automatically broadcast the original
transaction after a timeout of 1 minute, and only verify that it has
been propagated in the network.

However, non-interactive receivers (like a payment processor) need to
verify the transaction to prevent UTXO probing attacks.

This is not a concern for interactive receivers like Wasabi Wallet,
because those receivers can just limit the number of original PSBT
proposals of a specific address to one. With such wallets, the
attacker has no way to generate new deposit addresses to probe the
UTXOs.

===Spare change donation===

Small change inside wallets are detrimental to privacy. Mixers like
Wasabi wallet, because of its protocol, eventually generate such
[[https://docs.wasabiwallet.io/using-wasabi/ChangeCoins.html#first-round-coinjoin-change|small
change]].

A common way to protect your privacy is to donate those spare changes,
to deposit them in an exchange or on your favorite merchant's store
account. Those kind of transactions can easily be spotted on the
blockchain: There is only one output.

However, if you donate via payjoin, it will look like a normal transaction.

On top of this the receiver can poison analysis by randomly faking a
round amount of satoshi for the additional output.

===Payment output substitution===

The receiver is free to change the output paying to himself.
For example, if the sender's scriptPubKey type is P2WPKH while the
receiver's payment output in the original PSBT is P2SH, then the
receiver can substitute the payment output to be P2WPKH to match the
sender's scriptPubKey type.

===Impacted heuristics===

Our proposal of payjoin is breaking the following blockchain heuristics:

* Common inputs heuristics.

Because payjoin is mixing the inputs of the sender and receiver, this
heuristic becomes unreliable.

* Change identification from scriptPubKey type heuristics

When Alice pays Bob, if Alice is using P2SH but Bob's deposit address
is P2WPKH, the heuristic would assume that the P2SH output is the
change address of Alice.
This is now however a broken assumption, as the payjoin receiver has
the freedom to mislead analytics by purposefully changing the
invoice's address in the payjoin transaction.

Alternatively, if the original address of Bob is P2WPKH and Alice's
address is also P2WPKH, Bob can change the receiving address in the
payjoin to P2SH. The heuristic would wrongfully identify the payjoin's
receiving address as the change address of the transaction.

See payment output substitution above.

* Change identification from round change amount

If Alice pays Bob, she might be tempted to pay him a round amount,
like <code>1.23000000 BTC</code>. When this happens, blockchain
analysis often identifies the output without the round amount as the
change of the transaction.

For this reason, during a [spare
change](Payjoin-spec.md#spare-change-donation) situation, we randomly
round the amount in the output added by the receiver to the payjoin
transaction.

==Attack vectors==

===On the receiver side: UTXO probing attack===

When the receiver creates a payjoin proposal, they expose one or more
inputs belonging to them.

An attacker could create multiple original transactions in order to
learn the UTXOs of the receiver, while not broadcasting the payjoin
proposal.

While we cannot prevent this type of attack entirely, we implemented
the following mitigations:

* When the receiver detects an original transaction being broadcast,
or if the receiver detects that the original transaction has been
double spent, then they will reuse the UTXO that was exposed for the
next payjoin.
* While the exposed UTXO will be reused in priority to not leak other
UTXOs, there is no strong guarantee about it. This prevents the
attacker from detecting with certainty the next payjoin of the
merchant to another peer.

Note that probing attacks are only a problem for automated payment
systems such as BTCPay Server. End-user wallets with payjoin
capabilities are not affected, as the attacker can't create multiple
invoices to force the receiver to expose their UTXOs.

===On the sender side: Double payment risk for hardware wallets===

For a successful payjoin to happen, the sender needs to sign two
transactions double spending each other: The original transaction and
the payjoin proposal.

The sender's software wallet can verify that the payjoin proposal is
legitimate by the sender's checklist.

However, a hardware wallet can't verify that this is indeed the case.
This means that the security guarantee of the hardware wallet is
decreased. If the sender's software is compromised, the hardware
wallet would sign two valid transactions, thus sending two payments.

Without payjoin, the maximum amount of money that could be lost by a
compromised software is equal to one payment (via address
substitution).

With payjoin, the maximum amount of money that can be lost is equal to
two payments.

==Implementations==

* [[https://github.com/BlueWallet/BlueWallet|BlueWallet]] is in the
process of implementing the protocol.
* [[https://github.com/btcpayserver/btcpayserver|BTCPay Server]] has
implemented sender and receiver side of this protocol.
* [[https://github.com/zkSNACKs/WalletWasabi/|Wasabi Wallet]] has
merged sender's support.
* [[https://github.com/JoinMarket-Org/joinmarket-clientserver|Join
Market]] is in the process of implementing the protocol.
* [[https://github.com/junderw/payjoin-client-js|JavaScript sender
implementation]].

==Special thanks==

Special thanks to Kukks for developing the initial support to BTCPay
Server, to junderw, AdamISZ, lukechilds, ncoelho, nopara73, yahiheb
for all the feedback we received since our first implementation.
Thanks also to RHavar who wrote the
[[https://github.com/bitcoin/bips/blob/master/bip-0079.mediawiki|BIP79
Bustapay]] proposal, this gave a good starting point for our proposal.

[-- Attachment #2: Type: text/html, Size: 22942 bytes --]

                 reply	other threads:[~2020-05-16 19:46 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CA+1nnrkq2zq_PFoUyNadE7KMncbv30M=Kr-Ek4nDfnJvgDHrfQ@mail.gmail.com' \
    --to=nicolas.dorier@gmail.com \
    --cc=bitcoin-dev@lists.linuxfoundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox