Talos: Cellular Automata for Encryption

This is the code and RFC for a novel encryption algorithm based on cellular automata. We name the project after the mythical bronze automaton Talos, who defended the island of Crete from invaders.

Running the Project ⚒️

Compiling Talos

The rust implementation of our encryption algorithm can be built using cargo, installed via rustup. Once cargo has been installed, the project can be built with cargo build --release from the project root.

Encryption and Decryption

The current CLI tool for encryption is called crypt. To encrypt a file, one would use:

./crypt --encrypt --key <KEY> path/to/plain.txt -o encrypted.enc

crypt should automatically generate and print a key to stderr if one is not provided. Decryption is achievable with

./crypt --decrypt --key <KEY> path/to/encrypted.enc -o plain.txt

PyTorch Implementation

Additionally, we do provide a python implementation of the cellular automaton rule, although it is significantly slower than the rust implementation. The file, as well as the other python files in the script directory can be run after installing the dependencies in requirements.txt. I used uv to build my environment.

Compiling RFCs 📝

The RFCs contained in this repository can be compiled using typst.

Challenge ⚔️

An encrypted file has been provided via Git LFS. Users are free to attempt decryption via whatever means. Please let me know if you make any progress!

Disclosure and Warning ⚠️

It should be emphasized that this is merely a research exercise; I am not a crypanalyst or mathematician by trade, and using this tool for any serious endeavor would yield disasterous consequences.

Multiple steps which should be taken in the development of any cryptographic protocol have been intentionally ignored:

• No prior work in adapting CAs for encryption was consulted. This was done because I want to implement something unique, and did not want to be influenced by the state of the field.
• No large scale statistical tests have been employed to verify the pseudorandomness of our automata rule.
• Multiple weaknesses identified in the protocol have not been addressed; see RFCs for details.
• Time has not been taken to assess the algorithm deeply; this was put together in a week and a half because I found it interesting, and thought it would be a good learning opportunity.