Skip to content

alokmenghrajani/decv

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Deterministic ECDSA Cross Validation (DECV)

The purpose of DECV is to cross validate various different deterministic ECDSA implementations (libsecp256k1, OpenSSL, Trezor). By verifying that each library produces the exact same signatures for a large number of test vectors, we are able to confirm (with a high degree of confidence) that each library is both correct and lacks subliminal channels (also known as kleptograms). To learn more about ECDSA backdoors, see "Wallet Security" by Stephan Verbücheln and "Deterministic Signatures, Subliminal channels and Hardware wallets" by Sergio Demian Lerner.

We focus on curve secp256k1 since our application is signing Bitcoin transactions. Deterministic ECDSA is defined in rfc6979.

An ECDSA signature is represented as a pair of values (r, s). All implementations must generate the same r. However, there exists two valid values for s: s and -s mod n (where n is the order of the group). Another Bitcoin-centric decision is to always pick the lower s (see BIP: 62, BIP: 146).

python/decv.py is used to generate test vectors. These test vectors can be saved to a file or can be streamed to any other implementation. The test vectors also contain BIP32 derivations, which enables writing validation code which is as close as feasible to actual Bitcoin wallet code.

Note:

  • libsecp256k1 and OpenSSL are used via pycoin, a python library.
  • trezor is compiled using cmake, with build files copied from Subzero.

Running

The easiest way to run the code is to use Docker, as following. The code should run fine without Docker, as long as the various dependencies are available. It is recommended that you validate your crypto library in an environment identical to your production environment (i.e. it's preferable to validate hardware wallets on the actual hardware since the underlying library might run different code paths on different processors).

$ docker build -t decv . && docker run --rm -it decv
# ./python/decv.py --libsecp256k1 generate 10000 | ./python/decv.py --libsecp256k1 verify
verified: 10000 signatures
# ./python/decv.py --libsecp256k1 generate 10000 | ./python/decv.py --openssl verify
verified: 10000 signatures    
# ./python/decv.py --libsecp256k1 generate 10000 | ./trezor/build/decv
verified: 10000 signatures
# cat bip32_test_vectors.csv | ./python/decv.py --libsecp256k1 verify
verified: 14 signatures
# cat bip32_test_vectors.csv | ./python/decv.py --openssl verify
verified: 14 signatures
# cat bip32_test_vectors.csv | ./trezor/build/decv
verified: 14 signatures
# exit

Test vector format

The test vectors are emitted using comma separated values (CSV). See following table for the field names and one sample row.

seed (hex) chain ext pub (base58) ext priv (base58) hash of message (hex) signature (DER encoded, low s, hex)
78c72a6f7a2a488de34a11c1a7de6ab97133d321 m/0'/50/41/168'/115 xpub6FRofRU8HUx9T1cZLAvV46p7EsKL4TK4NXq2H2iVLn2CYBcTUaNbftBMZT9qqEnynndSZVVJhWwJKhER99Sa3Tjt5pS3CnBrrna4bhCNexV xprvA2STFuwET7PrEXY6E9PUgxsNgqUqezbD1JuRUeJsnSVDfPHJw34M85rsiBjERcEkunJ3kZ4N2Lg5xbbQ3UuastcwHVoF2H2ohpfSc4xV2GL bd7b0690546402a37af52e513bcdf965c15c4757e82354944552140727e08ede 304402201e924d772d72030794b25ea216298b57cbc09003f1db90691da1bfe1e292bd850220676f325432f9d14e7873cc64ea74ad88138c8e9a1793931a457523d7f7e4a79a

Cross Validated Implementations

Library Version
OpenSSL 1.1.1d
libsecp256k1 0.1~20170810-2
trezor-crypto e6d884b145d0fb6201c0ae76c552547028793df9

Releated and Future work

If you found DECV interesting or useful, you should check out Project Wycheproof and Cryptofuzz - Differential cryptography fuzzing. Both project, using different methodologies, look for bugs in cryptographic libraries.

At this point, DECV is considered complete and no future work is planned. However, we welcome pull requests which verify additional libraries or which verify existing libraries using different programming language wrappers.