Repository navigation
Conversation
One execution gives the paper's relaxed anonymous broadcast, where a corrupt client may choose its message after seeing the honest ones. The example shows the paper's lift to the full notion: two executions with fresh parameters, commitments first and openings second, matched at the recipient.
There was a problem hiding this comment.
🟡 Changes recommended
Duplicate commitment digests let a corrupt client copy an honest commitment and its revealed opening across executions, causing acceptance of an extra message dependent on the honest input and breaking input independence.
2 open issues
What changed in this PR
This PR documents the library’s relaxed anonymous-broadcast guarantee and illustrates application-side orchestration of two protocol executions, without changing library code.
Changes:
- Adds a README guarantee section explaining input dependence and the commitment/opening approach.
- Adds a two-run example with fresh parameters and server keys, cover traffic, a missing opening, and a mismatched opening.
| File | Description |
|---|---|
| README.md | Describes the relaxed guarantee and links to the two-run example. |
| examples/two_run.rs | Demonstrates commitment and opening rounds with recipient-side matching and output assertions. |
Review effort: Balanced
| let mut unopened: HashMap<[u8; COMMIT_BYTES], usize> = HashMap::new(); | ||
| for c in &commitments { | ||
| let c: [u8; COMMIT_BYTES] = c[..COMMIT_BYTES].try_into().expect("commitment width"); | ||
| *unopened.entry(c).or_default() += 1; |
noot
changed the title
Clarify the conditions for message delivery and the impact of malformed contributions. Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
|
Unfortunately the two-run version of the protocol is no longer a part of the paper. There was actually a bug in the security proof. The proof may or may not be fixable, however I found an easier and more efficient way of getting full security in a single run. That's what the current version of the paper describes. |
Adds a "Guarantee" section to the README and
examples/two_run.rs. No library change.paper divergence
The code implements one execution of the protocol, which gives the relaxed functionality from "Ideal Functionalities": honest clients stay anonymous, but the adversary sees the honest messages before the corrupt clients choose theirs, so a corrupt client's message may depend on them. "From Relaxed to Full Anonymous Broadcast" lifts this by running the protocol twice with fresh parameters and server keys, commitments first and openings second, matched at the recipient. That is orchestration over two ordinary executions, so it stays on the application side; the example shows the full pattern with a hash commitment, including a client that commits and never opens and one that opens to a different message. The README now states which guarantee the library delivers and points at the example.
The example calls
run_server_roundwith the signature on main; once the server verification branch merges it needs the two extra arguments.