This is a purely Rust-based library that provides both variants of the state-of-the-art Argon2 hashing algorithm, suitable for password hashing and password-based key derivation.
From examples/helloworld.rs:
extern crate argon2min;
pub fn main() {
let (password, salt) = ("argon2i!", "delicious salt");
println!("argon2i_simple(\"{}\", \"{}\"):", password, salt);
for byte in argon2min::argon2i_simple(&password, &salt).iter() {
print!("{:02x}", byte);
}
println!("");
}outputs:
argon2i_simple("argon2i!", "delicious salt"):
e254b28d820f26706a19309f1888cefd5d48d91384f35dc2e3fe75c3a8f665a6
There are two variants of Argon2 that differ in the manner by which reference indices are computed during block-filling rounds. Argon2d does this in a faster but data-dependent fashion that could be vulnerable to side-channel attacks, whereas Argon2i ("i" denoting independence from plaintext input) works slower but is immune to such attacks and is therefore the preferred choice for password hashing.
- Zero-on-drop trait for sensitive(s):
Matrix - Constant-time verification API.
- Benchmarks.
- Fuzz.
- Prove safety of unchecked accesses in
Block,Matrix.
Our primary benchmarks are single-threaded runs of Argon2i with default parameters against the reference implementation. In order to compile and run this, first pull in the C sources:
$ git submodule init
$ git submodule update benches/cargon/phc-winner-argon2and then benchmark with Cargo as usual:
$ export RUSTFLAGS='-C target-feature=+avx'
$ cargo bench --features="bench_ref"
# output trimmed for brevity
Running target/release/versus_cargon-b5955411e1594c85
running 5 tests
test ensure_identical_hashes ... ignored
test bench_argon2min_i ... bench: 6,856,774 ns/iter (+/- 197,405)
test bench_cargon_i ... bench: 3,856,783 ns/iter (+/- 142,580)
test result: ok. 0 passed; 0 failed; 0 ignored; 2 measured
"Argon2: The Memory-Hard Function for Password Hashing and Other Applications"