txoo

What is txoo?

txoo is an oracle for the Bitcoin network that produces attestations to the chain tip and to the outputs spent in each block. It uses GCS filters similar to BIP-158 compact block filters.

The oracle is useful for Lightning Network devices that need to verify the on-chain state of a channel without direct access to the Bitcoin network. This is especially useful for secure enclaves without network access.

no_std support is provided.

Security model

The oracle attests to the chain tip and to the outputs spent in each block at a specific time.

The outputs spent are summarized by a Golomb-coded set (GCS) filter representing the set of all outpoints in transaction inputs in the block. The attestation includes a "filter header", which hashes the current GCS filter as well as the previous filter header. This allows the client to verify the entire filter history has not been tampered with.

The proof of spend/non-spend includes the spending transactions, either in an SPV proof, or a full block if there is a filter false-positive.

Running txood

cargo run --package txood --release 

or if you have bitcoind running on the same machine, use this for the initial run, which will be much faster (~45 minutes on a fast machine):

cargo run --package txood --release -- --preload

check the last 256 attestations:

cargo run --package txoo-cli --release -- check

a quick local test starting at a recent block:

cargo run --package txood --release -- -d data --start-block 760000

Note that the filter headers won't match if you don't start at genesis, because it covers all previous blocks' filters.

Check an oracle

To check our testnet oracle, ensure bitcoind --testnet is running and:

cargo run -p txoo-cli -- -n testnet -s https://vls.tech/txoo/testnet/ check

Run coverage

cargo install cargo-kcov
./scripts/run-kcov

View Coverage Report

target/kcov/cov/index.html

Benchmark

cd txoo-bench
cargo bench

TODO

• emit attestations
• prune
• sign
• SPV proofs for spent outputs
• consider using multiple GCS filters with different sizes, so that we can have a small filter for applications that just watch a few outpoints, and a large filter for applications that watch a lot of outpoints