This is a utility that computes a Merkle tree hash over the specified files. The advantage that a Merkle tree has over applying the hash once includes
- Hash parallelism (which is not taken advantage of yet)
- Precisely identifying which blocks have been corrupted
The file is first divided into blocks that are block_length bytes long, and each block is hashed to create the leaf nodes of the tree. Adjacent nodes at each layer are grouped into groups of branch_factor nodes, and a hash is computed for a parent node (with these nodes as children) by hashing the concatenation of the hashes in the children nodes. (A parent node may have less than branch_factor children when the total number of blocks is not a power of branch_factor.)
To address second pre-image attacks, we prepend the byte 0x00 before hashing data blocks, and prepend 0x01 before hashing the concatenation of node hashes. (This is the same data adjustment used in Certificate Transparency.)