Philipp Krüger d24f960c53 feat: implement `PublicDirectory` conflict reconciliation (#426) | vor 7 Monaten | |
---|---|---|
.. | ||
proptest-regressions | vor 5 Monaten | |
src | vor 5 Monaten | |
CHANGELOG.md | vor 5 Monaten | |
Cargo.toml | vor 5 Monaten | |
README.md | vor 5 Monaten |
This library implements the cryptographic primitives necessary for WNFS to prove that its writes were valid in a way that's verifyable by third parties without read access.
Specifically, it implements 2048-bit RSA accumulators and the PoKE* and PoKCR algorithms from the paper "Batching Techniques for Accumulators with Applications to IOPs and Stateless Blockchains", as well as some WNFS-specific interfaces and serialized representations for them.
RSA accumulators require a trusted setup. Whoever has access to the trusted setup can create arbitrary valid proofs, which would in practice let malicious actors who've only been given partial access to a WNFS access to the rest of the file system. For this reason the trusted setup is run once upon creation of a new WNFS by the root author. The root author is naturally incentivized to throw away the toxic waste from the trusted setup.
use wnfs_nameaccumulator::{AccumulatorSetup, BatchedProofPart, BatchedProofVerification, Name, NameSegment};
use rand::thread_rng;
// Run the trutsed setup.
let rng = &mut thread_rng();
let setup = &AccumulatorSetup::trusted(rng);
// We want to prove the names for two files at
// /Docs/Note and /Pics/Image respectively
let mut name_note = Name::empty(setup);
let mut name_image = Name::empty(setup);
// Each segment is represented by a random 256-bit prime number
let root_dir_segment = NameSegment::new(rng);
let docs_dir_segment = NameSegment::new(rng);
let pics_dir_segment = NameSegment::new(rng);
let note_file_segment = NameSegment::new(rng);
let image_file_segment = NameSegment::new(rng);
name_note.add_segments([root_dir_segment.clone(), docs_dir_segment, note_file_segment]);
name_image.add_segments([root_dir_segment, pics_dir_segment, image_file_segment]);
// We can collapse these arrays of primes that represent paths into 2048-bit RSA accumulators
// with a proof that they were derived from the same "base" name, in this case the `Name::empty` above.
let (accum_note, proof_note) = name_note.as_proven_accumulator(setup);
let (accum_image, proof_image) = name_image.as_proven_accumulator(setup);
// Knowing the proofs, we can batch at least parts of the proofs together.
// This results in a single 2048-bit batched proof part and ~17-20 bytes of unbatched proof per element.
let mut batched_proof = BatchedProofPart::new();
batched_proof.add(&proof_note);
batched_proof.add(&proof_image);
// Without read access, but given the accumulated base name and the proofs,
// we can verify that the accumulated names are related to the same base name.
let name_base = Name::empty(setup).as_accumulator(setup).clone();
let mut verification = BatchedProofVerification::new(setup);
verification.add(&name_base, &accum_note, &proof_note.part)?;
verification.add(&name_base, &accum_image, &proof_image.part)?;
verification.verify(&batched_proof)?;
rug
featureThis enables a different backend for big unsigned integer arithmetic, based on the rug crate.
It is roughly 2x faster than the num-bigint-dig
implementation when building for release, but as a bonus is also fast in debug builds (e.g. during tests) due to rug containing a statically linked release build of GMP.
However, it doesn't work in Wasm and it should be noted that GMP is licensed as LGPLv3.
If you depend on the wnfs
crate, but want to use the rug
backend for your application, then simply add a wnfs-nameaccumulator
as a dependency and enable its rug
feature. This will make wnfs
use a version of wnfs-nameaccumulator
with rug
enabled:
wnfs-nameaccumulator = { version = "*", default-features = false, features = ["rug"] }