## DEV Community is a community of 665,273 amazing developers

We're a place where coders share, stay up-to-date and grow their careers.

# Implementing SHA2 (256/512) algorithm with Rust const generics

Alain Viguier
Senior software engineer & sysadmin, technical architect, Linux expert. Always willing to learn new stuff.

The const generics feature is landing in stable Rust from version 1.51. This feature has been longly-awaited, and you can get an idea of its possibilities here: Rust 1.51 release notes

I wanted to give it a try and thought that SHA2 message digest algorithms were a good test bed.

## SHA2 algorithm

I'm not going to dig into the details of the SHA message digest algorithm. It's a NIST standard (National Institute of Standards and Technology) labelled FIPS-180. You can get the detailed description in the NIST US government web site: FIPS-180-4

In a nutshell, following is the general algorithm:

• initialize the hash value and constants
• pad the massage to be a multiple of 64 or 128 (i.e. 512 or 1024 bits)
• for each block of data:
• prepare the message schedule using scrambling functions
• for each of the 64 or 80 rounds, compute intermediate hash value
• output hash value

I'm only dealing here with the hash of byte-oriented messages.

What can be noticed are the similarities between SHA224/256 and SHA384/512 in terms of functions, initial hash values,

Tables SHA224/256 SHA384/512
block size 64 bytes 128 bytes
padding 56 bytes boundary 112 bytes boundary
Initialization vector 8 4-byte values 8 8-byte values
K-constants 64 4-byte values 80 8-byte values
rounds 64 80
functions see below see below

## Rust implementation

Taking into account all those common features, we can define a SHA hash as:

``````// hash is either 256 or 512 bits but always 8 u32 or u64 integers
// T is either u32 or u64
pub struct Hash<T, const BLOCKSIZE: usize, const ROUNDS: usize> {
pub k_constants: [T; ROUNDS],   // ROUNDS = 64 or 80
pub hash: [T; 8],
pub scramble_funcs: ScramblePool<T>, // scrambling functions sigma etc
pub block: [u8; BLOCKSIZE], // BLOCKSIZE = 64 or 128
}
``````

### Scrambling functions

This terminology is not found in any NIST paper but I found it useful to name those functions. These are a combination of logical or, and, xor, bit rotation or bit shifting. But they are very similar:

• Ch(x,y,z) = (x & y) ^ (!x & z)
• Maj(x,y,z) = (x & y) ^ (x & z) ^ (y & z)
• Σ0 and Σ1 are the same, but with different shift values
• σ0 and σ1 are the same, but with different shift values

Using const generics, we can define those function generically, for u32 or u64 types but also for shifting values:

``````    // SIGMA(X) = RotR(X,A) ⊕ RotR(X,B) ⊕ RotR(X,C)
#[allow(non_snake_case)]
pub fn SIGMA<const A: u8, const B: u8, const C: u8>(x: T) -> T
where
T: BitAnd<Output = T>,
T: BitXor<Output = T>,
T: Shifter<T>,
T: Copy,
{
x.right_rotate(A) ^ x.right_rotate(B) ^ x.right_rotate(C)
}

// sigma(X) = RotR(X,A) ⊕ RotR(X,B) ⊕ X >> C
#[allow(non_snake_case)]
pub fn sigma<const A: u8, const B: u8, const C: u8>(x: T) -> T
where
T: Shr<Output = T>,
T: BitXor<Output = T>,
T: Shifter<T>,
T: Copy,
{
x.right_rotate(A) ^ x.right_rotate(B) ^ x.right_shift(C)
}
``````

where right_rotate() and right_shift() are trait functions implemented for u32 and u64.

The scrambling functions are then gathered into a generic structure:

``````// Rust doesn't allow yet utf-8 symbol for variables
#[allow(non_snake_case)]
pub struct ScramblePool<T> {
pub ch: FnScramble3<T>,
pub maj: FnScramble3<T>,
pub sigma0: FnScramble1<T>,
pub sigma1: FnScramble1<T>,
pub SIGMA0: FnScramble1<T>,
pub SIGMA1: FnScramble1<T>,
}
``````

### SHA2 algorithm

Because, of all this genericity, we can implement the SHA2 algorithm, bring either SHA256 or SHA512, totally generically. You can browse the code here: https://github.com/dandyvica/sha

### Monomorphization

Depending of which algorithm you want to use, it's easy to define an alias for a specific type:

``````pub type Sha256 = Hash<u32, 64, 64>;
pub type Sha512 = Hash<u64, 128, 80>;
``````

and implement the generic Hash for both u32 and u64:

``````// example for u32
impl Hash<u32, 64, 64> {
pub fn new() -> Self {
let iv: [u32; 8] = [
0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a, 0x510e527f, 0x9b05688c, 0x1f83d9ab,
0x5be0cd19,
];
Hash {
hash: iv,
k_constants: Self::k_constants(),
scramble_funcs: ScramblePool::<u32> {
ch: Scramble::<u32>::Ch,
maj: Scramble::<u32>::Maj,
sigma0: Scramble::<u32>::sigma::<7, 18, 3>,
sigma1: Scramble::<u32>::sigma::<17, 19, 10>,
SIGMA0: Scramble::<u32>::SIGMA::<2, 13, 22>,
SIGMA1: Scramble::<u32>::SIGMA::<6, 11, 25>,
},
block: [0u8; 64],
}
}

pub fn k_constants() -> [u32; 64] {
[
0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5, 0x3956c25b, 0x59f111f1, 0x923f82a4,
0xab1c5ed5, 0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3, 0x72be5d74, 0x80deb1fe,
0x9bdc06a7, 0xc19bf174, 0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc, 0x2de92c6f,
0x4a7484aa, 0x5cb0a9dc, 0x76f988da, 0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967, 0x27b70a85, 0x2e1b2138, 0x4d2c6dfc,
0x53380d13, 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85, 0xa2bfe8a1, 0xa81a664b,
0xc24b8b70, 0xc76c51a3, 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070, 0x19a4c116,
0x1e376c08, 0x2748774c, 0x34b0bcb5, 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208, 0x90befffa, 0xa4506ceb, 0xbef9a3f7,
0xc67178f2,
]
}
}

``````

### Testing with SHA2 test vectors

The NIST site give a list of files to be tested, for integration test. When running cargo t, those tests are also run and sha digests verified.

I didn't implement SHA224 or SHA384 but is left as an exercice 😀

Hope this help !

Photo by Conscious Design on Unsplash