This was the original intent of the sighash field. Either the sighash is acceptable to the signer and the signer signs with it, or they do not sign at all.

On 7/9/19 11:58 AM, Jonathan Underwood via bitcoin-dev wrote:
Hi all,

Just to be brief, I'll kick off with an attack scenario.

1. I am a signer, I get a PSBT that is ready to sign. I parse. I sign according to the PSBT as-is.
2. I notice my UTXO was stolen by a hacker because they changed my PSBT input's sighashtype to SIGHASH_ANYONECANPAY | SIGHASH_NONE and after the fact they changed the outputs to send to themselves, and added an input they signed with SIGHASH_ALL.
3. I lose the BTC in my UTXO.

So we should definitely add to the signer checks "ensure the sighash type given is the type of sighash you want to sign." etc.

My proposal for a wording change would be addition to the bullet list:

- If a sighash type is provided, the signer MUST check that the sighash type is acceptable to them, and fail signing if unacceptable.
- If a sighash type is not provided, the signer SHOULD sign using SIGHASH_ALL, but may sign with any sighash type they wish.

Any thoughts?

Thanks,
Jon

--
-----------------
Jonathan Underwood
ビットバンク社　チーフビットコインオフィサー
-----------------

暗号化したメッセージをお送りの方は下記の公開鍵をご利用下さい。

指紋: 0xCE5EA9476DE7D3E45EBC3FDAD998682F3590FEA3