Adding the hash Multimixer-128#591
Conversation
newpavlov
left a comment
newpavlov
left a comment
There was a problem hiding this comment.
AFAIK this hash function does not have any real world adoption and it has not received any independent cryptanalysis. The paper has only one citation from the same group. So I am not sure if we want to include this function at the current time.
@tarcieri WDYT?
| let mut k = [0u32; 4]; | ||
| for i in 0..4 { | ||
| h[i] = u32::from_ne_bytes([ | ||
| self.key_blocks[self.block_index][i * 4], |
There was a problem hiding this comment.
What happens with messages bigger than key size? IIUC this code will panic.
There was a problem hiding this comment.
Yes, correct, actually the message and key should be equal in size. I first created the algorithm with a key, but then also implemented a PRNG with a seed. I really wanted to bench the code, and for this I would need a really big key. Ideally I would maybe create two different hash-classes, Multimixer-128 and Multimixer-128-PRNG. I think this would solve the other comments as well.
| HashMarker, OutputSizeUser, | ||
| }; | ||
| use rand_chacha::rand_core::{RngCore, SeedableRng}; | ||
| use rand_chacha::ChaCha8Rng; |
There was a problem hiding this comment.
You should use the chacha20 crate instead.
| key_blocks: Vec<Block<Self>>, | ||
| block_sums: [u64; 8usize], | ||
| block_index: usize, | ||
| rng: Option<ChaCha8Rng>, |
There was a problem hiding this comment.
This looks... suspicious. If you want to make code generic over key blocks generator, I think it's better to make it generic over closure.
| } | ||
|
|
||
| //Uses key instead of RNG, needs to be same size as message. | ||
| fn new_from_slice(key: &[u8]) -> Result<Self, InvalidLength> { |
There was a problem hiding this comment.
In general, T::new(&key) and T::new_from_slice(&key).unwrap() should produce the same result for the same key.
|
Thank you for the review! Yes, it should not be used by the public, but as I understood it, maybe as a replacement for the NU hash in UMAC. I really appreciate the comments, but if it not going to be merged, I think I'll give implementing UMAC a try 🙂. RustCrypto/MACs#1 |
|
Joan Daemen is listed as one of the authors on the paper, so I wouldn't completely discount it, but I'm not seeing a whole lot in the way of peer review on it. It seems very new. |
3 participants
Did this to exercise my Rust. Any interest having it in the main repo?