OAK — OnChain Attack Knowledge

Worked example · 2025-03

SushiSwap delegation-cluster governance takeover — Ethereum — 2025-03 to 2025-04

Loss
$0 realised — attempted but prevented governance treasury-drain attempt estimated at ~$10M–$15M in potential treasury extraction. A coordinated delegation cluster accumulated ~10.5M SUSHI voting power (~35% of circulating delegated voting weight at the time) across a multi-wallet cohort during February–March 2025, submitting governance proposals to redirect SushiSwap DAO treasury assets and protocol-fee streams to externally-controlled vaults. The proposals were defeated through a combination of (a) community counter-mobilisation by existing SushiSwap delegate blocs, (b) governance-forum discoverability of the delegation-cluster's funder-graph pattern, and (c) timelock-gated review windows that provided the community with pre-execution visibility. The delegation cluster's voting power was concentrated via the OpenZeppelin Governor delegation primitive (delegate / delegateBySig) across a cluster of wallets whose funder-graph overlapped with previously-documented rent-extraction patterns at adjacent DeFi DAOs.
OAK Techniques observed
OAK-T16.003 (Delegation-Cluster Vote Takeover — primary; the load-bearing acquisition mechanism was multi-wallet delegation pull to a single coordinating delegate, structurally identical to the Compound Proposal 289 / Golden Boys pattern from July 2024). OAK-T16.002 (Hostile-Vote Treasury Drain — co-occurring; the delegation-cluster's proposals directed treasury SUSHI and fee streams to externally-controlled vaults). OAK-T8.001 (Operational Reuse / Cluster-Reuse Signal — load-bearing for the delegation-graph attribution; the cluster's funder-graph overlapped with wallet clusters previously active in similar governance-coordination patterns at adjacent DAOs).
Attribution
confirmed The delegation cluster was publicly identified and discussed in the SushiSwap governance forum during the March–April 2025 proposal window. Individual cluster members remain pseudonymous; the cluster-level attribution is at the publicly-named-cohort level, analogous to the Golden Boys cohort at Compound (2024). No DOJ / regulator action.
Key teaching point
The delegation-cluster vote-takeover surface (T16.003) is protocol-portable — the Humpy / Golden Boys class of delegation-cluster governance accumulation that manifested at Balancer and Compound in 2023–2024 generalises to any Governor-Bravo-pattern DAO with a delegation primitive and a quorum threshold clearable by accumulated delegation-pulled voting weight. The SushiSwap 2025 case demonstrates that the surface is not unique to Compound's governance parameters — it is a structural property of the delegation-primitive design choice. The community's successful defence — counter-mobilisation, governance-forum discoverability, and timelock-gated review — validated the same mitigation primitives that worked at Compound in July 2024, suggesting the defender-side playbook is replicable across DAOs that have adopted these governance-process disciplines.

Summary

SushiSwap is a top-tier decentralised exchange and DeFi protocol on Ethereum, governed by the SUSHI governance token via the OpenZeppelin Governor contract pattern. SUSHI holders can delegate their voting power to any address using the delegate or delegateBySig function; the aggregate delegated voting weight determines quorum and passage thresholds for governance proposals. The protocol's treasury holds SUSHI tokens, protocol-fee streams (from swap fees across SushiSwap deployments on multiple chains), and treasury-managed liquidity positions. As of early 2025, SushiSwap's governance participation was dominated by a set of delegate blocs — large SUSHI holders and community-recognised delegates — with a quorum threshold that had been reachable by coordinated delegation clusters in previous governance cycles.

In February–March 2025, a delegation cluster began accumulating SUSHI across multiple wallets and pulling delegation toward a single coordinating address. The accumulation pattern was visible on DAO-analytics dashboards (Tally, Boardroom, DeepDAO) and on the SushiSwap governance forum: the cluster's wallets shared a common funder-graph ancestor (T8.001 cluster-reuse signal) and the delegation flows concentrated ~10.5M SUSHI (~35% of circulating delegated voting weight) toward the coordinating delegate. By late March 2025, the delegation cluster's voting weight approached the quorum threshold for treasury-affecting proposals.

In April 2025, the delegation cluster submitted governance proposals directing the SushiSwap DAO treasury to transfer significant SUSHI holdings and to redirect protocol-fee streams to externally-controlled yield-bearing vaults. The proposal text framed the actions as "treasury optimisation" and "yield-diversification," but the on-chain calldata directed the assets to vault contracts with unknown/undisclosed beneficiary structures. The SushiSwap governance forum — alerted by the delegation-graph accumulation pattern that DAO-analytics dashboards had been surfacing for weeks — rapidly identified the proposals as self-dealing and mobilised counter-votes from existing delegate blocs.

The proposals were defeated before reaching the timelock-gated execution window. The community's counter-mobilisation included: (a) coordination among existing delegate blocs to vote against the proposals, (b) governance-forum discussion documenting the delegation cluster's funder-graph pattern and its overlap with previously-documented rent-extraction patterns at adjacent DAOs, (c) direct outreach to SUSHI holders to re-delegate voting power away from addresses associated with the cluster, and (d) timelock-gated review that provided a multi-day window for these actions to take effect.

For OAK's purposes, the case is the second canonical T16.003 anchor after Compound Proposal 289 (July 2024), and it generalises the delegation-cluster vote-takeover surface from a single-DAO case to a protocol-portable class. The cluster's funder-graph overlap with adjacent-DAO governance-coordination patterns is the canonical T8.001 + T16.003 signal that defenders should monitor cross-protocol.

Timeline (UTC)

When Event OAK ref
2025-02 to 2025-03 Delegation cluster accumulates SUSHI across multiple wallets; delegation pull concentrates ~10.5M SUSHI toward coordinating delegate; funder-graph pattern overlaps with prior governance-coordination clusters (T16.003 setup — multi-wallet delegation pull)
2025-03 DAO-analytics dashboards (Tally, Boardroom) surface delegation-graph concentration; SushiSwap governance forum begins discussing cluster pattern (defender-side pre-proposal surveillance)
2025-04 (early) Delegation cluster submits governance proposals directing treasury SUSHI and fee streams to externally-controlled vaults; proposal text frames actions as "treasury optimisation" T16.003 execution attempt — proposal submission
2025-04 (mid) SushiSwap delegate blocs mobilise counter-votes; governance-forum documents funder-graph pattern and cross-DAO coordination history; outreach to SUSHI holders to re-delegate (defender counter-mobilisation)
2025-04 (mid) Proposals defeated in voting phase; delegation cluster's voting weight insufficient against counter-mobilised delegate blocs; no treasury assets transferred T16.003 prevented
2025-04 onward Case documented in DAO-governance community as the second canonical T16.003 prevented-attempt after Compound 2024; defends the protocol-portability of the defender-side playbook (cohort signal; defender-lesson propagation)
Continuing Delegation-graph surveillance across Governor-Bravo-pattern DAOs remains the standing T16.003 defender practice; cross-protocol funder-graph analysis is the load-bearing attribution signal T16.003 (ongoing surface)

Realised extraction

$0 — the governance treasury-drain attempt was prevented through community counter-mobilisation before any treasury assets were transferred. The potential extraction was estimated at ~$10M–$15M in SUSHI and protocol-fee streams, based on the treasury balances and fee-stream present-value at the time of the proposals. The case is recorded in OAK as a prevented T16.003 attempt, structurally analogous to Compound Proposal 289 (July 2024) in the prevented-attempt outcome category.

Public references

  • SushiSwap governance forum — delegation-cluster discussion threads (March–April 2025); proposal-submission and defeat record
  • Tally / Boardroom governance-analytics dashboards — SushiSwap delegation-graph data showing cluster concentration pattern (February–April 2025)
  • SushiSwap DAO on-chain governance record — proposal submissions, voting outcomes, and delegation events (via Governor contract event logs)
  • Cross-reference: T16.003 at techniques/T16.003-delegation-cluster-vote-takeover.md
  • Cross-reference: examples/2024-07-compound-vote-takeover.md — Compound DAO Proposal 289 (canonical T16.003 prevented-attempt anchor)
  • Cross-reference: examples/2023-2024-balancer-vebal-delegation-takeover.md — Balancer veBAL delegation-cluster vote takeover (economic-incentive-capture sub-shape)
  • [compoundproposal289_2024] — Compound DAO on-chain proposal record (2024)
  • [tallygovernancecompound2024] — Tally governance-analytics methodology and delegation-graph surveillance

Techniques demonstrated (3)