Hacktoberfest 2026：メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

P3 dispute: upheld S4 dispute slashes colluders but never replaces the wrong finalCID or refunds the honest dissenter

オープン
  • 何か問題がありますか？
#194 コメント 1 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
solidity
領域
blockchain, security

調査の方向性

Read foundry/src/DINTaskCoordinator.sol around settleRecomputation and the bad consensus slash in DINTaskCoordinator.sol:971, then review Developer/design/adversarial-threat-model.md row 5 and Part 2's test_knownGap_t1WrongCID. First define how confirmed disputes recover dissenters' slashes and install a recomputed CID through downstream T2 and global model state; completion should make test_defended_t1WrongCID pass.

索引モデルが issue の本文から書いたものです。

説明

enhancement

Summary

An upheld and confirmed S4 dispute punishes the colluding T1 aggregators but does not repair the damage. The wrong CID stays the batch's result, and the honest dissenter stays slashed.

Scenario: 2 of 3 aggregators in a T1 batch collude and submit the same wrong CID. Majority vote finalizes it, and the honest third aggregator is slashed a full minStake for AGG_T1_BAD_CONSENSUS (foundry/src/DINTaskCoordinator.sol:971), because its CID disagrees with the (wrong) majority.

The S4 path is openDispute → lockDisputeSeed → resolveDispute(upheld=true) → settleRecomputation(confirmed=true). On confirmed=true, settleRecomputation (DINTaskCoordinator.sol:1351-1400):

  • returns the challenger's bond (disputeBondClaimable),
  • slashes the aggregators whose submission matched the disputed CID, S4_INVALID_AGGREGATION,
  • does not replace tier1Batches[_GI][batchId].finalCID (or the T2 equivalent), so the wrong model remains in that GI's aggregation,
  • does not refund the honest dissenter's AGG_T1_BAD_CONSENSUS slash.

_assignFreshSubgroup is bookkeeping only and does not re-open the GI state machine. The fresh subgroup never submits a CID on-chain, so there is no corrected CID to install even if the contract wanted to.

Impact

This holds even with an honest owner-adjudicator (TA-1). The attack partially succeeds: the wrong model propagates for that GI, and the only validator who behaved correctly loses a full minStake. That also discourages honest dissent in future batches.

This is separate from No. 181, which is about who adjudicates disputes (removing the owner-adjudicator trust assumption). This issue is about what a successful dispute does, whoever adjudicates it.

Needed

  1. Refund the dissenter: on confirmed=true, reverse the AGG_T1_BAD_CONSENSUS / AGG_T2_BAD_CONSENSUS slash for batch members whose submission differed from the disputed finalCID. This needs either a refund path on DinValidatorStake or the bad-consensus slash held pending until the dispute window closes.
  2. Repair the result: have the fresh subgroup submit its recomputed CID on-chain and install it as finalCID on confirmed=true, or define what happens downstream (T2 / next GI's global model) when a finalized CID is later shown to be wrong.

Both touch slashing/dispute mechanism design, so this likely needs a design note before implementation.

Tracking

Row 5 (KNOWN GAP, decided in PR No. 182's review) of the adversarial threat model, Developer/design/adversarial-threat-model.md (task_240926_17). Part 2's test_knownGap_t1WrongCID should flip to test_defended_t1WrongCID when this is fixed.

主要言語
Python
スター
4
フォーク
7
平均マージ
3日 58分
マージ済み PR（30日）
39

環境構築

このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

InfiniteZeroFoundation/DevNet のほかの issue

InfiniteZeroFoundation/DevNet の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。