Cyclone-Hash

A modular, state-interdependent hash algorithm with tweak-based diffusion.

Usage:

The python script can be run through the terminal with python3 cyclone.py "message" "tweak". If no tweak is provided, one will be generated.

You can also import the script and use the provided hash() function. The function uses the following parameters:

message     : Mandatory, the message you want to hash
chunkLength : The length of chunk to use in bytes. Essentially, this is the length of the output hash in bytes. Default is 32 bytes.
tweak       : The tweak to use with the mixer. If you don't provide one, one will be generated for you.

If you just want to quickly hash a message, you can run hash("Hello, World"), and a 32 byte (256 bit) hash generated with a random tweak will be returned.

Security Test Stats:

Hamming distance between small changes (“Hello, World!” vs “Hfllo, World!” vs “Hello, Wprld!”) averages ~49%
In a trial running “msg0000” to “msg9999”, the changes in the hash bytes were almost fully evenly distributed. The amount of changes (per byte) were [9962, 9964, 9956, 9957, 9953, 9965, 9958, 9971, 9965, 9963, 9949, 9961, 9964, 9952, 9965, 9960]
With multiple birthday-paradox-influenced collision detection tests with randomly generated input strings, no collisions have been found

For reference: the optimal scores for each test are the following:

Hamming distance: 50%.
Byte Diffusion: graphs of the amount of changes per byte should be flat or almost flat.
Birthday paradox collisions: 0 collisions.

Tests were performed with a constant tweak value, which indicates that the use of a tweak value increases the strength of the hash beyond the test results above.

Note: the tests were applied to a 16 byte (128 bit) long hash to increase testing efficiency. The default value is 32 bytes, or 256 bits. Hamming tests and diffusion tests for 32 bytes have similar results (Hamming: ~50.3%, Diffusion: ~99.5% average, or ~9950 average changes)

Background

Originally, I did this to give myself an introduction to hash algorithms and how they work internally, but through multiple hours of wondering what I could do better and the occasional eureka moment, my “introductory exercise” became Cyclone.

Developed independently by Samuel Joseph Poyner (Blooper7) and licensed under the MIT license

This site is open source. Improve this page.