horcrux: Split files into encrypted fragments for Node & React Native (TypeScript library)

#typescript #reactnative #node #showdev

Horcrux is a TypeScript library for splitting files into encrypted fragments — “horcruxes” — inspired by Harry Potter. Think of it as digital horcruxes for your sensitive files. You can split your files into multiple pieces where only a subset is needed to restore the original — no password required.

The original idea comes from jesseduffield/horcrux — the CLI tool from the lazygit author that slices a file into N shares and restores it when any M shares are present.

I wanted a version that plugs into both Node and React Native projects, so “Yet Another Horcrux” became a library.

Repo: https://github.com/nao1215/horcrux

npm: https://www.npmjs.com/package/@nao1215/horcrux

How to use

The API stays simple: use split and bind for files, splitBuffer and bindHorcruxes if you’re juggling data in memory. Both sides run on Node and React Native.

Use files

import { split, bind, nodeAdapter } from '@nao1215/horcrux';
import { saveHorcruxes } from '@nao1215/horcrux/core/split';

// Split a file into 5 pieces, need 3 to restore
const result = await split('secret.pdf', 5, 3);

// Save horcruxes (creates secret_1_of_5.horcrux, etc.)
const files = await saveHorcruxes(result.horcruxes, './output', nodeAdapter);

// Later: Restore from any 3 horcruxes
await bind(files.slice(0, 3), 'restored_secret.pdf');

Use buffer

import { splitBuffer } from '@nao1215/horcrux';
import { bindHorcruxes } from '@nao1215/horcrux/core/bind';

const data = Buffer.from('Secret message');
const result = await splitBuffer(data, 'message.txt', {
  total: 4,
  threshold: 2
});

const restored = await bindHorcruxes(result.horcruxes.slice(0, 2));
console.log(restored.data.toString()); // 'Secret message'

How it works

Under the hood it’s Shamir’s Secret Sharing (SSS).

First the file is encrypted, then the encryption key is encoded as coefficients of a polynomial. Each share gets the ciphertext plus a polynomial evaluation. Set a threshold t and total n; as soon as t shares show up, Lagrange interpolation reconstructs the polynomial and the key drops out.

Sit at t-1 and you’re stuck with infinite candidates — guessing isn’t happening. Each share isn’t just “n equal slices”; it carries redundant structure to make the threshold trick work while staying useless alone.

Why did I make this?

In Japan, there’s an internet meme: “If I die, destroy my HDD (or SSD).”
My counterpoint: “I’ve split my files with horcrux, so unless you know the restoration procedure, you can’t see what’s inside — no need to destroy anything, right?”