[bug]: Remote Signer: FundPsbt: change output lacking master fingerprint (XFP)

[kaloudis]

Background

In trying to craft a transaction using an input from an external watch-only account, it appears that the change output lacks a master fingerprint (XFP). The input has the XFP marked correctly.

This causes problems when signing with a Coldcard - error message: Change Fraud. Output#1: Deception regarding change output. BIP-32 path doesn't match actual address

Your environment

• lnd v0.17.4-beta
• Mobile: Android + iOS
• Neutrino backend
• Coldcard Q1

PSBT in question:

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

Steps to reproduce

• Import xpub from external account using ImportAccount endpoint. Make sure master fingerprint (XFP) is defined in the process.
• Deposit funds to external account
• Craft transaction using FundPsbt using input from the external account
• Attempt to sign PSBT with Coldcard

Expected behaviour

Change output has defined XFP that matches the wallet.

Actual behaviour

Change output has null XFP

[guggero]

Thanks for the report. It looks like we indeed forgot to set the master fingerprint in some situations on newly derived addresses. Was able to reproduce and fix locally with #8630.
Can you confirm please?

[kaloudis]

The XFP (master fingerprint) is now being added correctly. Thanks Oli!

We have another issue now though. It seems that we're hitting a mismatch on the pubkey.

Pubkey 0303b42b33356ebae804c43c2910666d5b6a4c141e86f501d4a8d597a27a91b796 != the result of (m=C13E4B30)/84'/0'/0'/1/5

PSBT in question:

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

[kaloudis]

It seems we also have an incorrect path for the input of 0h/0h/0/0 instead of 1h/0h/0/0 - testnet path not being taken into account here?

[guggero]

Ah, I know what this is... Unfortunately, the btcwallet doesn't use the coin type for testnet/regtest/signet at all for the default address types (BIP-44, 49, 84, 86), only for the lnd internal keys (purpose 1017').
Can you test if this works correctly when you derive the watch-only account from m/84'/0'/0' directly? Or does the cold card then complain about wanting testnet?

[kaloudis]

Ah, I know what this is... Unfortunately, the btcwallet doesn't use the coin type for testnet/regtest/signet at all for the default address types (BIP-44, 49, 84, 86), only for the lnd internal keys (purpose 1017'). Can you test if this works correctly when you derive the watch-only account from m/84'/0'/0' directly? Or does the cold card then complain about wanting testnet?

I think it needs to be in testnet mode, otherwise the addresses are calculated incorrectly.

Seems like we need to fix the derivation path used when importing tpubs via the ImportAccount API? Derivation path still shows as m/84'/0'/0' with a tpub

I will try to do some testing on mainnet to see if the issue persists there

[kaloudis]

Hitting the same issue when using a testnet seed with Seedsigner.

[guggero]

Yeah, it probably won't work with any hardware wallet, since btcwallet internally uses the wrong derivation path for testnet. Did you try on mainnet, since you gave a tACK on the PR?
I don't think we can just fix the testnet derivation calculation in btcwallet without also breaking a lot of other things on testnet...

[kaloudis]

No problem on mainnet! I have tested it and was able to sign with the hardware wallet.

I created a separate issue to track the testnet derivation path issue: #8645