shake128

Shake128

This is a pure (safe) Rust implementation of shake128. This code should be easy to adapt to all SHA-3 hash family since the general Keccak1600 has been implemented in the code.

Build

cd code

# Ensure that everything is clean and test runs
cargo clippy --all -- -D warnings
cargo test

# Build
RUSTFLAGS="-C target-cpu=native" cargo build --release

# Copy to PWD
cp target/release/shake128 ../
cd ../

Run

Please first execute the build steps.

./shake128 [HASH SIZE (in bytes)] < /path/to/input/file

Benchmark

Summary

Note : generated using hyperfine on a ~78Mo file, see below the table for the full output

Implementation	Time consumed (absolute)
OpenSSL	198 ms
Python (hashlib)	209 ms
Rust (tiny-keccak, quoted in Keccak website)	198 ms
My implementation	208 ms

Output :

Benchmark 1: ./target/release/shake128 4096 < /initrd.img.old
  Time (mean ± σ):     208.0 ms ±   1.6 ms    [User: 187.3 ms, System: 20.7 ms]
  Range (min … max):   206.0 ms … 211.4 ms    14 runs
 
Benchmark 2: openssl dgst -shake128 -xoflen 4096 < /initrd.img.old
  Time (mean ± σ):     198.3 ms ±   2.3 ms    [User: 184.9 ms, System: 13.3 ms]
  Range (min … max):   196.3 ms … 204.2 ms    15 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Benchmark 3: python3 -c 'import hashlib; import sys; print(hashlib.shake_128(sys.stdin.buffer.read()).hexdigest(4096))' < /initrd.img.old
  Time (mean ± σ):     208.9 ms ±   1.7 ms    [User: 193.5 ms, System: 15.3 ms]
  Range (min … max):   206.8 ms … 212.4 ms    14 runs
 
Benchmark 4: ../keccak_ref/target/release/keccak_ref 8192 < /initrd.img.old
  Time (mean ± σ):     203.4 ms ±   1.0 ms    [User: 186.1 ms, System: 17.2 ms]
  Range (min … max):   202.1 ms … 205.3 ms    14 runs
 
Summary
  openssl dgst -shake128 -xoflen 4096 < /initrd.img.old ran
    1.03 ± 0.01 times faster than ../keccak_ref/target/release/keccak_ref 4096 < /initrd.img.old
    1.05 ± 0.01 times faster than ./target/release/shake128 4096 < /initrd.img.old
    1.05 ± 0.01 times faster than python3 -c 'import hashlib; import sys; print(hashlib.shake_128(sys.stdin.buffer.read()).hexdigest(4096))' < /initrd.img.old

Methodology

I tried to do a fair comparison, so every programm reads the input from the STDIN. However, the python programm should be a bit faster because of the just-in-time compilation that induces some overhead. Moreover, the two rust programms are compiled using the -C target-cpu=native flag (openssl is surely compiled with equivalent flags, this it seems fair for a benchmark).