Technical help needed regarding Electrum mnemonics

Coding Enthusiast

legendary

Activity: 1042

Merit: 2805

Bitcoin and C♯ Enthusiast

Yeah I do realize that! But the problem isn't the HMAC function, it instead is what's being passed to this function.

You see unlike BIP-39, Electrum heavily modifies the words in the given string before it passes them to the HMAC function. For example in the case of the test vector linked above all spaces are being removed then it is passed to the HMAC function as "msg". That makes the data part 35 bytes whereas it normally would have somewhere around 80+.

Code:

眼悲叛改节跃衡响疆股遂冬 -> 眼悲叛改节跃衡响疆股遂冬

Basically my code was simply missing these 2 methods:
https://github.com/Autarkysoft/Denovo/blob/648419d8a3ccd590051ab390539b0b7147917d4e/Src/Autarkysoft.Bitcoin/ImprovementProposals/ElectrumMnemonic.cs#L234-L258

Abdussamad

legendary

Activity: 3682

Merit: 1580

see here:

https://en.wikipedia.org/wiki/HMAC

in electrum it's being used as a fancy hash function. the message is the seed mnemonic and the key is simply the string "seed version" in byte form.

Coding Enthusiast

legendary

Activity: 1042

Merit: 2805

Bitcoin and C♯ Enthusiast

Could you also tell me what "msg" does it pass this function:
https://github.com/spesmilo/electrum/blob/8d7370d897314d8542906aecc6a45cc949651f77/electrum/crypto.py#L299

Abdussamad

legendary

Activity: 3682

Merit: 1580

it's outputting the following when using the functions in electrum:

Code:

1002a06ca7de987ae74c4189cfdc3cf45bb3a836d21019bc19e265d2858e3eccc7b26c9a5509a4f7ccb9f9fc96663d268ca2a1e49dd2d08af832e96fccaf95c5

so it's a valid segwit seed.

Coding Enthusiast

legendary

Activity: 1042

Merit: 2805

Bitcoin and C♯ Enthusiast

I'm trying to add Electrum mnemonic recovery option to The FinderOuter and looking at the test_mnemonic.py file I'm having trouble understanding the tests even though I think I understand how the mnemonic.py file works.

The English test vectors in this file and also random keys I created using Electrum work fine meaning after computing HmachSha512(data=words, key="Seed version") I get a digest that starts with SEED_PREFIX (01) or SEED_PREFIX_SW (100). eg. the first digest is 1001bc7d1ea.... which is all Electrum looks for in a valid mnemonic (there is no checksum).

But the rest don't.
Take the Chinese case for example, the digest is

Code:

0f5c4c9ff66e87bcbdde59f06ad540eba48fe06f7bdfa16b1248cb868ae3cd3fe7f6ea08e50bc77fdec1f08d8b5710bd25a9d76427e636feed23cbcb3ec8b8cb

Which is incorrect.
It is worth mentioning that the words are normalized using form KD as is with BIP-39 and return the same bytes as the "words_hex" in test vector.

So what is the problem here?

Topic: Technical help needed regarding Electrum mnemonics (Read 206 times)