WringTwistree.jl

Whole-message cipher and tree hash, Julia implementation
Author phma
Popularity
0 Stars
Updated Last
6 Months Ago
Started In
December 2023

WringTwistree

Wring is a whole-message cipher. This is like a block cipher, but the block can be arbitrarily large, up to about 3 GiB. Wring can also be used as a length-preserving all-or-nothing transform.

Twistree is a hash function. It constructs two trees out of the blocks of data, runs the compression function at each node of the trees, and combines the results with the compression function.

Both algorithms are keyed. The key is turned into 96 16-bit words, which are then used to make three S-boxes. The key can be any byte string, including the empty string.

Features

A round consists of four operations:

  1. The mix3Parts operation splits the message or block in three equal parts and mixes them nonlinearly. This provides both diffusion and nonlinearity. The number of rounds in Wring increases logarithmically with message size so that diffusion spreads to the entire message.
  2. The three key-dependent 8×8 S-boxes provide confusion and resist linear cryptanalysis.
  3. Rotating by the population count thwarts integral and differential cryptanalysis by moving the difference around to a different part of the message.
  4. Wring's round constant, which is dependent on the byte position as well as the round number, is added to every byte to prevent slide attacks and ensure that an all-zero message doesn't stay that way.
  5. Twistree runs a CRC backwards to make the four bytes about to be dropped affect all the other bytes.

Julia

This is a Julia implementation. The reference implementations, in Rust and Haskell, are in the wring-twistree repo.

In the project directory, run julia --project at the shell prompt, then in Julia run using WringTwistree. You probably want export JULIA_NUM_THREADS=auto in your .profile, as the algorithms run faster on big inputs when multithreaded. You can then create a Wring, a Twistree, and some byte vectors and encrypt and hash them.

Before encrypting, decrypting, or hashing, and after upgrading Julia, run setBreakEven(). This sets the break-even points for Wring and Twistree. When given more data than the break-even points, the functions will handle them in parallel.

Test vectors

Test vectors are in test/runtests.jl. To run them, type ] to enter Pkg mode, then test.

Used By Packages

No packages found.