* v2.1.0.2 * restore tweaks to simlibp2p * fix: nil ref on size calc * fix: panic should induce shutdown from event_distributor * fix: friendlier initialization that requires less manual kickstarting for test/devnets * fix: fewer available shards than provers should choose shard length * fix: update stored worker registry, improve logging for debug mode * fix: shut the fuck up, peer log * qol: log value should be snake cased * fix:non-archive snap sync issues * fix: separate X448/Decaf448 signed keys, add onion key to registry * fix: overflow arithmetic on frame number comparison * fix: worker registration should be idempotent if inputs are same, otherwise permit updated records * fix: remove global prover state from size calculation * fix: divide by zero case * fix: eager prover * fix: broadcast listener default * qol: diagnostic data for peer authenticator * fix: master/worker connectivity issue in sparse networks tight coupling of peer and workers can sometimes interfere if mesh is sparse, so give workers a pseudoidentity but publish messages with the proper peer key * fix: reorder steps of join creation * fix: join verify frame source + ensure domain is properly padded (unnecessary but good for consistency) * fix: add delegate to protobuf <-> reified join conversion * fix: preempt prover from planning with no workers * fix: use the unallocated workers to generate a proof * qol: underflow causes join fail in first ten frames on test/devnets * qol: small logging tweaks for easier log correlation in debug mode * qol: use fisher-yates shuffle to ensure prover allocations are evenly distributed when scores are equal * qol: separate decisional logic on post-enrollment confirmation into consensus engine, proposer, and worker manager where relevant, refactor out scoring * reuse shard descriptors for both join planning and confirm/reject decisions * fix: add missing interface method and amend test blossomsub to use new peer id basis * fix: only check allocations if they exist * fix: pomw mint proof data needs to be hierarchically under global intrinsic domain * staging temporary state under diagnostics * fix: first phase of distributed lock refactoring * fix: compute intrinsic locking * fix: hypergraph intrinsic locking * fix: token intrinsic locking * fix: update execution engines to support new locking model * fix: adjust tests with new execution shape * fix: weave in lock/unlock semantics to liveness provider * fix lock fallthrough, add missing allocation update * qol: additional logging for diagnostics, also testnet/devnet handling for confirmations * fix: establish grace period on halt scenario to permit recovery * fix: support test/devnet defaults for coverage scenarios * fix: nil ref on consensus halts for non-archive nodes * fix: remove unnecessary prefix from prover ref * add test coverage for fork choice behaviors and replay – once passing, blocker (2) is resolved * fix: no fork replay on repeat for non-archive nodes, snap now behaves correctly * rollup of pre-liveness check lock interactions * ahead of tests, get the protobuf/metrics-related changes out so teams can prepare * add test coverage for distributed lock behaviors – once passing, blocker (3) is resolved * fix: blocker (3) * Dev docs improvements (#445) * Make install deps script more robust * Improve testing instructions * Worker node should stop upon OS SIGINT/SIGTERM signal (#447) * move pebble close to Stop() * move deferred Stop() to Start() * add core id to worker stop log message * create done os signal channel and stop worker upon message to it --------- Co-authored-by: Cassandra Heart <7929478+CassOnMars@users.noreply.github.com> --------- Co-authored-by: Daz <daz_the_corgi@proton.me> Co-authored-by: Black Swan <3999712+blacks1ne@users.noreply.github.com> |
||
|---|---|---|
| .. | ||
| benches | ||
| src | ||
| build.rs | ||
| Cargo.lock | ||
| Cargo.toml | ||
| README.md | ||
RPM
Based on the paper RPM: Robust Anonymity at Scale. Built to be used as a primitive – all sharp edges are exposed, using this without properly checking the boundaries of the input will cut!
Includes UniFFI UDL file for FFI integration.
How to use safely
There are six public functions of interest for builders to use if they want to have a purpose built mixnet:
- rpm_generate_initial_shares
- rpm_combine_shares_and_mask
- rpm_sketch_propose
- rpm_sketch_verify
- rpm_permute
- rpm_finalize
rpm_generate_initial_shares
Performs the first of four steps of the "offline" phase of RPM Variant 3 (Subvariant 2). Each active dealer in a given offline batch generates initial shares of the permutation matrices and masks. Assumes player identifiers are in monotonically increasing order starting from 1, player count must be greater than or equal to the square of dealers. For malicious security, a ZKPoK should be used for each share set and broadcasted by all dealers prior to the next step.
rpm_combine_shares_and_mask
Performs the second of four steps of the "offline" phase. Each player should have received their respective shares and organized them according to sequence of players for inputs. For malicious security, the ZKPoK should be verified after invoking this. The splits of the permutation matrices and masks should be passed into the next step.
rpm_sketch_propose
Performs the third of four steps of the "offline" phase. Each player should take the splits of the permutation matrices and masks and pass them into this function. All players should broadcast the sketch proposals.
rpm_sketch_verify
Performs the fourth step of the "offline" phase. Each player should take the broadcasted sketch proposals and pass them into the function. If any dealer or player had cheated, it would be revealed by the intersection of failures.
rpm_permute
Performs the first step of the "online" phase. Before invoking this, each message sender should obtain a vector of the first depth's mask shares, at the same index from each, from as many players as needed to match the dealer count. The sender should combine these shares to produce the mask for their message, and add it to their message (chunked by field element size if need be), and secret share it with the same parameters (t = dealers, n = players). The players will collect their respective shares of inputs into an input vector following the order of the mask shares that were applied. The players will then pass in the input vector shares, matrix shares, mask shares, and combined matrix/mask shares from rpm_combine_shares_and_mask, along with the depth and player identifiers in player order. This function should be invoked in as many rounds as there is depth to the mixnet, using the previous invocation's output shares given to each respective player. This must be repeated for each chunked vector of field elements if applicable.
rpm_finalize
A convenience funciton which recombines the shares of the final invocation of rpm_permute's output into the finalized vector of the mixnet.