refactored mnemonic.py

lbry/wallet/mnemonic.py

@@ -1,159 +1,57 @@
-# Copyright (C) 2014 Thomas Voegtlin
-# Copyright (C) 2018 LBRY Inc.
-
-import hmac
-import math
 import hashlib
-import importlib
+import asyncio
 import unicodedata
-import string
 from binascii import hexlify
-from secrets import randbelow
-
-import pbkdf2
+from secrets import randbits
 
 from lbry.crypto.hash import hmac_sha512
-from .words import english
-
-# The hash of the mnemonic seed must begin with this
-SEED_PREFIX = b'01'       # Standard wallet
-SEED_PREFIX_2FA = b'101'  # Two-factor authentication
-SEED_PREFIX_SW = b'100'   # Segwit wallet
-
-# http://www.asahi-net.or.jp/~ax2s-kmtn/ref/unicode/e_asia.html
-CJK_INTERVALS = [
-    (0x4E00, 0x9FFF, 'CJK Unified Ideographs'),
-    (0x3400, 0x4DBF, 'CJK Unified Ideographs Extension A'),
-    (0x20000, 0x2A6DF, 'CJK Unified Ideographs Extension B'),
-    (0x2A700, 0x2B73F, 'CJK Unified Ideographs Extension C'),
-    (0x2B740, 0x2B81F, 'CJK Unified Ideographs Extension D'),
-    (0xF900, 0xFAFF, 'CJK Compatibility Ideographs'),
-    (0x2F800, 0x2FA1D, 'CJK Compatibility Ideographs Supplement'),
-    (0x3190, 0x319F, 'Kanbun'),
-    (0x2E80, 0x2EFF, 'CJK Radicals Supplement'),
-    (0x2F00, 0x2FDF, 'CJK Radicals'),
-    (0x31C0, 0x31EF, 'CJK Strokes'),
-    (0x2FF0, 0x2FFF, 'Ideographic Description Characters'),
-    (0xE0100, 0xE01EF, 'Variation Selectors Supplement'),
-    (0x3100, 0x312F, 'Bopomofo'),
-    (0x31A0, 0x31BF, 'Bopomofo Extended'),
-    (0xFF00, 0xFFEF, 'Halfwidth and Fullwidth Forms'),
-    (0x3040, 0x309F, 'Hiragana'),
-    (0x30A0, 0x30FF, 'Katakana'),
-    (0x31F0, 0x31FF, 'Katakana Phonetic Extensions'),
-    (0x1B000, 0x1B0FF, 'Kana Supplement'),
-    (0xAC00, 0xD7AF, 'Hangul Syllables'),
-    (0x1100, 0x11FF, 'Hangul Jamo'),
-    (0xA960, 0xA97F, 'Hangul Jamo Extended A'),
-    (0xD7B0, 0xD7FF, 'Hangul Jamo Extended B'),
-    (0x3130, 0x318F, 'Hangul Compatibility Jamo'),
-    (0xA4D0, 0xA4FF, 'Lisu'),
-    (0x16F00, 0x16F9F, 'Miao'),
-    (0xA000, 0xA48F, 'Yi Syllables'),
-    (0xA490, 0xA4CF, 'Yi Radicals'),
-]
+from . import words
 
 
-def is_cjk(c):
-    n = ord(c)
-    for start, end, _ in CJK_INTERVALS:
-        if start <= n <= end:
-            return True
+def get_languages():
+    return words.languages
+
+
+def normalize(mnemonic: str) -> str:
+    return ' '.join(unicodedata.normalize('NFKD', mnemonic).lower().split())
+
+
+def is_valid(language, mnemonic):
+    local_words = getattr(words, language)
+    for word in normalize(mnemonic).split():
+        if word not in local_words:
             return False
+    return bool(mnemonic)
 
 
-def normalize_text(seed):
-    seed = unicodedata.normalize('NFKD', seed)
-    seed = seed.lower()
-    # remove accents
-    seed = ''.join([c for c in seed if not unicodedata.combining(c)])
-    # normalize whitespaces
-    seed = ' '.join(seed.split())
-    # remove whitespaces between CJK
-    seed = ''.join([
-        seed[i] for i in range(len(seed))
-        if not (seed[i] in string.whitespace and is_cjk(seed[i-1]) and is_cjk(seed[i+1]))
-    ])
-    return seed
-
-
-def load_words(language_name):
-    if language_name == 'english':
-        return english.words
-    language_module = importlib.import_module('lbry.wallet.client.words.'+language_name)
-    return list(map(
-        lambda s: unicodedata.normalize('NFKD', s),
-        language_module.words
-    ))
-
-
-LANGUAGE_NAMES = {
-    'en': 'english',
-    'es': 'spanish',
-    'ja': 'japanese',
-    'pt': 'portuguese',
-    'zh': 'chinese_simplified'
-}
-
-
-class Mnemonic:
-    # Seed derivation no longer follows BIP39
-    # Mnemonic phrase uses a hash based checksum, instead of a words-dependent checksum
-
-    def __init__(self, lang='en'):
-        language_name = LANGUAGE_NAMES.get(lang, 'english')
-        self.words = load_words(language_name)
-
-    @staticmethod
-    def mnemonic_to_seed(mnemonic, passphrase=''):
-        pbkdf2_rounds = 2048
-        mnemonic = normalize_text(mnemonic)
-        passphrase = normalize_text(passphrase)
-        return pbkdf2.PBKDF2(
-            mnemonic, passphrase, iterations=pbkdf2_rounds, macmodule=hmac, digestmodule=hashlib.sha512
-        ).read(64)
-
-    def mnemonic_encode(self, i):
-        n = len(self.words)
-        words = []
-        while i:
-            x = i%n
-            i = i//n
-            words.append(self.words[x])
-        return ' '.join(words)
-
-    def mnemonic_decode(self, seed):
-        n = len(self.words)
-        words = seed.split()
-        i = 0
-        while words:
-            word = words.pop()
-            k = self.words.index(word)
-            i = i*n + k
-        return i
-
-    def make_seed(self, prefix=SEED_PREFIX, num_bits=132):
-        # increase num_bits in order to obtain a uniform distribution for the last word
-        bpw = math.log(len(self.words), 2)
-        # rounding
-        n = int(math.ceil(num_bits/bpw) * bpw)
-        entropy = 1
-        while 0 < entropy < pow(2, n - bpw):
-            # try again if seed would not contain enough words
-            entropy = randbelow(pow(2, n))
+def sync_generate(language: str) -> str:
+    local_words = getattr(words, language)
+    entropy = randbits(132)
     nonce = 0
     while True:
         nonce += 1
         i = entropy + nonce
-            seed = self.mnemonic_encode(i)
-            if i != self.mnemonic_decode(seed):
-                raise Exception('Cannot extract same entropy from mnemonic!')
-            if is_new_seed(seed, prefix):
+        w = []
+        while i:
+            w.append(local_words[i % 2048])
+            i //= 2048
+        seed = ' '.join(w)
+        if hexlify(hmac_sha512(b"Seed version", seed.encode())).startswith(b"01"):
             break
     return seed
 
 
-def is_new_seed(seed, prefix):
-    seed = normalize_text(seed)
-    seed_hash = hexlify(hmac_sha512(b"Seed version", seed.encode('utf8')))
-    return seed_hash.startswith(prefix)
+def sync_to_seed(mnemonic: str) -> bytes:
+    return hashlib.pbkdf2_hmac('sha512', normalize(mnemonic).encode(), b'lbryum', 2048)
+
+
+async def generate(language: str) -> str:
+    return await asyncio.get_running_loop().run_in_executor(
+        None, sync_generate, language
+    )
+
+
+async def to_seed(mnemonic: str) -> bytes:
+    return await asyncio.get_running_loop().run_in_executor(
+        None, sync_to_seed, mnemonic
+    )

lbry/wallet/words/__init__.py

@@ -0,0 +1,8 @@
+from .english import words as en
+from .french import words as fr
+from .italian import words as it
+from .japanese import words as ja
+from .portuguese import words as pt
+from .spanish import words as es
+from .chinese_simplified import words as zh
+languages = 'en', 'fr', 'it', 'ja', 'pt', 'es', 'zh

tests/unit/wallet/test_mnemonic.py

@@ -1,23 +1,42 @@
-import unittest
+from unittest import TestCase
 from binascii import hexlify
 
-from lbry.wallet.mnemonic import Mnemonic
+from lbry.wallet import words
+from lbry.wallet.mnemonic import (
+    get_languages, is_valid,
+    sync_generate as generate,
+    sync_to_seed as to_seed
+)
 
 
-class TestMnemonic(unittest.TestCase):
+class TestMnemonic(TestCase):
 
-    def test_mnemonic_to_seed(self):
-        seed = Mnemonic.mnemonic_to_seed(mnemonic='foobar', passphrase='torba')
+    def test_get_languages(self):
+        languages = get_languages()
+        self.assertEqual(len(languages), 6)
+        for lang in languages:
+            self.assertEqual(len(getattr(words, lang)), 2048)
+
+    def test_is_valid(self):
+        self.assertFalse(is_valid('en', ''))
+        self.assertFalse(is_valid('en', 'foo'))
+        self.assertFalse(is_valid('en', 'awesomeball'))
+        self.assertTrue(is_valid('en', 'awesome ball'))
+
+        # check normalize works (these are not the same)
+        self.assertTrue(is_valid('ja', 'るいじ りんご'))
+        self.assertTrue(is_valid('ja', 'るいじ りんご'))
+
+    def test_generate(self):
+        self.assertGreaterEqual(len(generate('en').split()), 11)
+        self.assertGreaterEqual(len(generate('ja').split()), 11)
+
+    def test_to_seed(self):
         self.assertEqual(
-            hexlify(seed),
-            b'475a419db4e991cab14f08bde2d357e52b3e7241f72c6d8a2f92782367feeee9f403dc6a37c26a3f02ab9'
-            b'dec7f5063161eb139cea00da64cd77fba2f07c49ddc'
+            hexlify(to_seed(
+                "carbon smart garage balance margin twelve che"
+                "st sword toast envelope bottom stomach absent"
+            )),
+            b'919455c9f65198c3b0f8a2a656f13bd0ecc436abfabcb6a2a1f063affbccb628'
+            b'230200066117a30b1aa3aec2800ddbd3bf405f088dd7c98ba4f25f58d47e1baf'
         )
-
-    def test_make_seed_decode_encode(self):
-        iters = 10
-        m = Mnemonic('en')
-        for _ in range(iters):
-            seed = m.make_seed()
-            i = m.mnemonic_decode(seed)
-            self.assertEqual(m.mnemonic_encode(i), seed)