bip39

An implementation of BIP39 mnemonics for deterministic key generation.

Usage

A sample GHCi session:

  > -- pragma and bip32/TIO imports not required; just for illustration
  > :set -XOverloadedStrings
  > import qualified Crypto.HDKey.BIP32 as BIP32
  > import qualified Data.Text.IO as TIO
  >
  > import Crypto.KDF.BIP39
  >
  > -- bring your own entropy, e.g. via System.Entropy
  > -- for now we'll use an ultra-secret source
  > let trop = "my ultra secret entropy!"
  >
  > -- use 'mnemonic' to create a BIP39 mnemonic
  > let mnem = mnemonic trop
  > mnem
  "hope simple bubble suggest elbow correct limb hole gloom nasty fringe dolphin finger demand situate unlock junior autumn"
  >
  > -- you can use other wordlists; the bip39 'defaults' are included
  > -- to supply an alternative one, use '_mnemonic' (with a leading underscore)
  > let fr_mnem = _mnemonic french trop
  > let jp_mnem = _mnemonic japanese trop
  >
  > TIO.putStrLn fr_mnem
  frivole reculer bilingue secouer diriger chose ingérer frémir fatal luxueux étonnant défiler éolien crayon réflexe tournage hélium appareil
  >
  > TIO.putStrLn jp_mnem
  ずあん はんらん えんえん へいしゃ けむし きかんしゃ そめる しんせいじ したて ちきゅう さとし けいたい こんだて ぐうせい ひかく もちろん せんげん いやがる
  >
  > -- use 'seed' to create a seed
  > let master_seed = seed mnem mempty
  >
  > -- the second argument is a passphrase, which can be arbitrary UTF-8
  > let phrase_seed = seed jp_mnem "㍍ガバヴァぱばぐゞちぢ十人十色"
  >
  > -- use a seed to derive a bip32 hd wallet
  > let hd_wallet = BIP32.master phrase_seed

Documentation

Haddocks (API documentation, etc.) are hosted at docs.ppad.tech/bip39.

Security

This library aims at the maximum security achievable in a garbage-collected language under an optimizing compiler such as GHC, in which strict constant-timeness can be challenging to achieve.

The implementation within passes the official BIP39 test vectors for every "default" wordlist, including the additional Japanese vectors with heavily-normalized passphrases.

This library differs somewhat from most ppad libraries as it includes a somewhat-nonstandard dependency, namely text-icu, which wraps the International Components for Unicode library that is required for UTF8 normalization.

text-icu itself is venerable, is maintained by the Haskell Libraries Committee on GitHub, and its FFI code is simple and easy to audit. The alternatives to including it would probably be either only to support English wordlists, or to vendor the icu4c dependency ourselves and create our own FFI bindings to the NFKD normalization code we care about. The latter may be something I look to do in a future release.

If you discover any vulnerabilities, please disclose them via security@ppad.tech.

Performance

The aim is best-in-class performance for pure, highly-auditable Haskell code.

Current benchmark figures on my mid-2020 MacBook Air look like (use cabal bench to run the benchmark suite):

  benchmarking ppad-bip39/mnemonic
  time                 4.222 μs   (4.055 μs .. 4.380 μs)
                       0.992 R²   (0.990 R² .. 0.995 R²)
  mean                 4.220 μs   (4.118 μs .. 4.332 μs)
  std dev              366.1 ns   (318.3 ns .. 427.9 ns)
  variance introduced by outliers: 84% (severely inflated)

  benchmarking ppad-bip39/seed
  time                 12.91 ms   (12.64 ms .. 13.35 ms)
                       0.994 R²   (0.989 R² .. 0.997 R²)
  mean                 12.87 ms   (12.62 ms .. 13.16 ms)
  std dev              699.8 μs   (568.8 μs .. 938.4 μs)
  variance introduced by outliers: 25% (moderately inflated)

Development

You'll require Nix with flake support enabled. Enter a development shell with:

$ nix develop

Then do e.g.:

$ cabal repl ppad-bip39

to get a REPL for the main library.