#chunks #arweave #packing #decryption #block #data #entropy

arweave_rs_packing

Contains the packing and feistel decryption methods for unpacking chunked Arweave data

1 unstable release

0.1.0 Feb 14, 2024

#4 in #arweave


Used in arweave_rs_validator

MIT/Apache

51KB
779 lines

Library package that contains functions related to packing/unpacking arweave chunks. Presently used in validating the poa and poa2 chunks included in Arweave block headers.

Packing

The main packing function is

pub fn compute_entropy(
    input: &[u8],
    randomx_program_count: usize,
    randomx_vm: Option<&RandomXVM>,
) -> [u8; RANDOMX_ENTROPY_SIZE] {
  ...
}

Which takes a byte buffer and performs RandomX hashing on it to generate a scratchpad. Currently RANDOMX_ENTROPY_SIZE is 256Kib, the size of an Arweave chunk.

the input for each chunks entropy comes form the following consensus method.

pub fn get_chunk_entropy_input(
    chunk_offset: U256,
    tx_root: &H256,
    reward_addr: &H256
) -> [u8; 32] {
    let mut chunk_offset_bytes: [u8; 32] = [0; 32];
    chunk_offset.to_big_endian(&mut chunk_offset_bytes);

    let mut hasher = sha::Sha256::new();
    hasher.update(&chunk_offset_bytes);
    hasher.update(tx_root.as_bytes());
    hasher.update(reward_addr.as_bytes());
    hasher.finish()
}

The chunk_offset is the absolute offset of the chunk in the weave data set, and the tx_root comes from the block which contains the chunk. Currently tx_root is retrieved from the BlockIndex during validation.

Decrypting

Arweave chunks are packed using a combination of the original 256KiB bytes of the chunk combined with 256KiB bytes of RandomX entropy. To pack the chunk the original bytes of the chunk and the entropy are combined using a feistel block cypher.To get the original bytes out of a packed chunk means generating the randomX entropy for the packed chunk and using that as in input to feistel decript the original chunk bytes.

Dependencies

~8–18MB
~254K SLoC