A pure Haskell implementation of SHA-512 and HMAC-SHA512 on strict and lazy ByteStrings, as specified by RFC's 6234 and 2104, that uses ARM SHA2 or x86 SHA512 intrinsics when available.
A sample GHCi session:
> :set -XOverloadedStrings
>
> -- import qualified
> import qualified Crypto.Hash.SHA512 as SHA512
>
> -- 'hash' and 'hmac' operate on strict bytestrings
>
> let hash_s = SHA512.hash "strict bytestring input"
> let hmac_s = SHA512.hmac "strict secret" "strict bytestring input"
>
> -- 'hash_lazy' and 'hmac_lazy' operate on lazy bytestrings
> -- but note that the key for HMAC is always strict
>
> let hash_l = SHA512.hash_lazy "lazy bytestring input"
> let hmac_l = SHA512.hmac_lazy "strict secret" "lazy bytestring input"
>
> -- results are always unformatted 512-bit (64-byte) strict bytestrings
>
> import qualified Data.ByteString as BS
>
> BS.take 10 hash_s
"\189D*\v\166\245N\216\&1\243"
> BS.take 10 hmac_l
"#}9\185\179\233[&\246\205"
>
> -- you can use third-party libraries for rendering if needed
> -- e.g., using base64-bytestring:
>
> import qualified Data.ByteString.Base64 as B64
>
> B64.encode (BS.take 16 hash_s)
"vUQqC6b1Ttgx8+ydx4MmtQ=="
> B64.encode (BS.take 16 hmac_l)
"I305ubPpWyb2zUi4pwDkrw=="
Haddocks (API documentation, etc.) are hosted at docs.ppad.tech/sha512.
The aim is best-in-class performance. We avail of hardware
acceleration via the ARM SHA512 extensions on aarch64, and the x86
SHA512 extensions on x86_64, whenever the CPU supports them. Current
benchmark figures on an M4 Silicon MacBook Air look like (use cabal bench to run the benchmark suite):
benchmarking ppad-sha512/SHA512 (32B input)/hash
time 129.7 ns (129.1 ns .. 130.3 ns)
1.000 R² (1.000 R² .. 1.000 R²)
mean 130.1 ns (129.9 ns .. 130.3 ns)
std dev 668.1 ps (432.4 ps .. 974.3 ps)
benchmarking ppad-sha512/HMAC-SHA512 (32B input)/hmac
time 359.6 ns (356.2 ns .. 365.1 ns)
0.998 R² (0.994 R² .. 1.000 R²)
mean 360.3 ns (356.3 ns .. 371.5 ns)
std dev 18.82 ns (2.196 ns .. 34.70 ns)
variance introduced by outliers: 70% (severely inflated)
You should compile with the 'llvm' flag for maximum performance.
The security properties of this library have been examined and defended in a security analysis at ppad.tech.
The HMAC-SHA512 functions within pass all Wycheproof vectors, as well as various other useful unit test vectors found around the internet.
If you discover any vulnerabilities, please disclose them via security@ppad.tech.
You'll require Nix with flake support enabled. Enter a development shell with:
$ nix develop
Then do e.g.:
$ cabal repl ppad-sha512
to get a REPL for the main library.
This implementation has benefitted immensely from the SHA package available on Hackage, which was used as a reference during development. Many parts wound up being direct translations.