Swell Liquid Staking

Tier 3

Security Analysis Report

TVL

$28.50M

Sections

7

report sections

Citations

358

evidence references

Version

v1

10/29/2026

5.6

Moderate Risk

Safety Score · out of 10

📋

Executive Summary

5.6/10

Executive Summary: Swell Liquid Staking

Swell Liquid Staking (the swETH product) is in an open incident. DefiLlama shows about $28.5M of TVL (deposits), all on Ethereum, and the sources do not say whether that figure covers swETH, rswETH or both . A Swell X post dated 2026-10-23 (source reliability: unverified publisher) announces a pause of deposits and withdrawals for swETH and rswETH. The pause flags were still set on-chain on 2026-10-29. The cause, the size of any rate error and any loss are not confirmed by a reliable source. Treat the position as impaired until Swell publishes a corrected rate, a root cause and a reopening plan.

Safety Rating

Overall Score: 5.6/10 Dimension scores: Code security 6.0 · Economic security 3.9 · Team transparency 5.7 · Governance 5.0 · Track record 8.0

Scores by dimension, points out of 10 (points)
Code security
6.0 points
Economic security
3.9 points
Team transparency
5.7 points
Governance
5.0 points
Track record
8.0 points

The rating reflects an unresolved incident, a closed exit route, and audits that cannot be tied to the contracts running today. Scores below are shown as code-rendered values; this text does not restate them.

Scores by Dimension

DimensionKey Finding
Code SecurityOne Hexens review of the 2024 swETH update found 1 medium and 7 low/informational findings, but no audit is tied to a deployed address
Economic SecurityWithdrawals are paused, so principal is inaccessible through the protocol and solvency cannot be shown
Team TransparencySwell has a named legal operator, but only one named leader, and that profile rests on third-party pages
GovernanceOn the measured rswETH path, a small multisig behind a 7-day timelock, not token holders, controls upgrades; fund-moving roles and swETH controls are unresolved or not measured, and captures of the Snapshot voting space (unverified publishers) show no proposals ; on the measured rswETH side one pauser signer can freeze exits
Track RecordListed since 2023-04-27 with no DefiLlama hacks record, but an October 2026 pause of unconfirmed cause and a reported April 2026 bridge pause (from Swell X posts, source reliability: unverified publisher, unconfirmed) sit in the history

The Track Record card is mechanical: the cross-analysis notes it uses an empty hacks dataset and "does not reflect the live event" .

What This Protocol Does

Swell lets a user stake ETH without running a validator and receive swETH, a token whose value against ETH rises as validators earn rewards. ETH is pooled into groups of 32 ETH, and each group becomes a validator run by a permissioned node operator. The record covers swETH only. rswETH (liquid restaking) and other family products are context, except where they share code, keys or incident exposure.

Documented vs On-Chain

  • Swell's docs describe the rate as set in part by a Chainlink proof-of-reserve oracle with a daily balance update . The on-chain rswETH contract also has a reprice function with a 1% cap per step and a one-hour minimum . The overview concludes the rate is likely moved by a privileged bot-role submitter. The swETH reprice holder and cap were not read.
  • A Swell X post of 2026-10-23 (source reliability: unverified publisher) says deposits and withdrawals are paused. On Ethereum at block 26300483 (2026-10-29), the swETH access-control contract still reports withdrawals paused.
  • A press report (source reliability: unverified publisher) puts the last clean swETH rate near 1.133 ETH. On-chain at block 26300482 (2026-10-29) the contract rate is about 1.203 ETH. The clean rate is unconfirmed.
  • An archived Balancer review (source reliability: unverified publisher) lists the admin multisig as 3/5. On-chain at block 26300146 (2026-10-29) the rswETH admin Safe is 4-of-7, and the older value is treated as stale.

Key Strengths

  • Upgrades on the measured rswETH contracts pass a 7-day timelock administered by a 4-of-7 Safe. For swETH the delay is inferred from a shared owner address (an assessment, not a direct read) . The governance report does not confirm swETH owners or delays and calls the match to rswETH unverified .
  • swETH (2024 liquid staking update) had an independent Hexens review: 1 medium and 7 low/informational findings, all fixed or acknowledged. Its match to deployed code is not confirmed.
  • Withdrawals go through a queue, so a same-transaction deposit-and-redeem loop is blocked by design.
  • Swell publishes a named legal operator, contract address tables and audit lists.

Key Risks

  • Exits stay closed and the rate may be reset (magnitude: swETH and rswETH exit route closed; likelihood: high that exits stay closed until Swell acts). A Swell X post (source reliability: unverified publisher, unconfirmed) says the next step is to return NAV (the exchange rate) to the correct value before reopening. The size and direction of the correction are unconfirmed. If the press-reported clean rate (unverified publisher) were right, the swETH rate would fall about 5.8%. For rswETH a single signer of a 1-of-5 Safe can pause, and, for swETH only, the audited 2024 update had a pause with no automatic resume (acknowledged); whether the current swETH deployment differs was not compared, and the sources do not show that rswETH shares that behaviour .
  • Misuse of the rate-reporting (bot) role after reopening (magnitude: unknown for swETH, severe for rswETH if repeated capped steps compound; likelihood: medium, an assessment with low confidence). Who holds the swETH repricing role is not known, and no swETH step cap is confirmed. The audited design checks the bot's figure against a proof-of-reserves oracle, but that is the audited version and not shown to be current.
  • Compromised signers or infrastructure reaching the upgrade path (magnitude: loss of pooled ETH reachable by upgraded contracts; likelihood: low). A press report (source reliability: unverified publisher) alleges that four malicious upgrade contracts able to take about 2,500 ETH were deployed but not queued. Both the deployment and the amount are unconfirmed, and the cross-analysis calls the related arithmetic an unconfirmed input . Nothing ties any signer to a breach. The 7-day timelock covers only the measured rswETH upgrade path: the delay and holders for the rswETH staker-proxy upgrade function and other fund-moving permissions are unresolved, so the timelock may not cover every path to funds . The claim that nothing was queued comes only from the press report (source reliability: unverified publisher), and the timelock queue was not read directly . The low band rests on the 7-day public window on the measured path .
  • No audit tied to the contracts running today. Deployed code was not compared with any audited version, so this is "not compared", not a measured mismatch. The latest swETH audit is from February 2024, before the October 2026 incident.
  • Thin exit liquidity. The only measured DEX depth (rswETH) is zero at 2% price impact, and swETH depth is not measured.

Critical Findings

  • The Hexens 2024 liquid staking review's single medium finding (token-queue reentrancy by a BOT-role holder) was fixed.
  • The code report lists two critical items, both from the 2022-03-11 SoftStack/Chainsulting audit of an earlier swETH version. One (swNFTUpgrade.sol, initialize not protected) is acknowledged; the team says it is guarded at proxy level, and this was not checked against code. The other is reported in the audit's executive summary, not itemized, and marked fixed.
  • Operationally, the open October 2026 pause is the most urgent matter. Its cause and any loss are unconfirmed.

Recommendation

Not suitable for users who need access to their ETH soon or who cannot tolerate a possible exchange-rate reset. Withdrawals are closed, the corrected rate is unpublished, and the key facts (who holds the swETH rate and pause roles, and backing versus liabilities) are not public. Only risk-tolerant holders who can wait should stay exposed. Do not assume 1.203 ETH per swETH on redemption. New deposits should wait for a post-mortem, a published corrected rate and key rotation notice.

Report Sections

  • Protocol Overview
  • Code Security
  • Economic Security
  • Social & Team
  • Governance
  • Cross-Risk Analysis

39 citations · click markers to view evidence

🏗

Protocol Overview

Protocol Overview: Swell Liquid Staking

Conclusion. Swell Liquid Staking is the swETH product: users deposit ETH and receive swETH, a token whose value against ETH rises as validators earn rewards. The DefiLlama record shows about $28.5M on Ethereum as of 2026-10-29. As of that date the product is in an unresolved incident. Swell paused swETH and rswETH deposits and withdrawals on 2026-10-23, and on-chain pause flags were still set on 2026-10-29. The swETH contract rate was about 1.203 ETH that day, which press reports describe as inflated. Treat the product as impaired until Swell confirms the exchange rate is corrected and reopens redemptions. The cause of the incident has not been confirmed by Swell.

What It Does

Swell lets a user stake ETH (lock it to help secure Ethereum and earn rewards) without running a validator or having 32 ETH. In return the user gets swETH, a "liquid" token that can be held or used in DeFi while the underlying ETH keeps earning. Swell describes itself as a non-custodial staking protocol: .

  • Category: liquid staking. The DefiLlama category is .
  • Product scope of this record: swETH, the ETH liquid staking token, on Ethereum. The record is one member of a larger Swell family. The other DeFiLlama members are supersweth-vault, swell-btc-lrt, swell-earn, swell-l2-farm and swell-liquid-restaking. Those are other products. rswETH (liquid restaking) is covered here only where it shares code, keys or incident exposure with swETH.
  • Target users: ETH holders who want staking yield and a token usable elsewhere in DeFi.
  • Related token: SWELL, the governance token, has a maximum supply of 10,000,000,000 . It is not part of the staking yield.

How It Works

The core design is pooled staking. ETH from many users is pooled into groups of 32 ETH, and each group becomes a validator run by a permissioned node operator. A validator is the Ethereum node that attests to blocks and earns rewards.

  1. The user stakes ETH into the swETH contract, and swETH is minted to the user.
  2. The swETH contract sends the ETH to the Deposit manager contract, which pools it until at least 32 ETH is available.
  3. The next validator in the node operator registry is picked round robin, and the deposit goes to the Ethereum deposit contract with that validator's key.
  4. The validator waits in the consensus-layer activation queue, then attests and proposes blocks.
  5. Rewards raise the swETH-to-ETH exchange rate; the balance in the user's wallet stays the same.
  6. To exit, a holder either withdraws through the Swell app at the primary market rate or swaps swETH on a DEX. Swell's docs give typical waits of 1-7 days, with occasional delays of up to 12 days or more . That page is marked as last updated one year before the snapshot, and it cites a sweep delay of up to 8.5 days "as of March '24" . Each withdrawal request is represented by an NFT, minted when the request is made .
swETH stake flow: ETH in, validator outOpen full view →
ACTORETH stakerCONTRACTswETH contractCONTRACTDeposit managercontractCONTRACTNode operator registryEXTERNALEthereum depositcontractstakes ETH,receives swETH1sends ETH to bepooled2selects nextvalidator3deposits 32 ETHper validator3
  • (actor)
  • (contract)
  • (contract)
  • (contract)
  • (external)
  • step 1flow
  • step 2flow
  • step 3flow
  • step 3flow

Key contracts (Ethereum, from Swell's address list). The list gives these for swETH: swETH token, swETH Deposit Manager and swETH Node Operator Registry . Their addresses are in the source.

  • swETH (0xf951E335afb289353dc249e82926178EaC7DEd78): token and exchange-rate holder.
  • swETH Deposit Manager (0xb3D9cf8E163bbc840195a97E81F8A34E295B8f39): pools ETH and makes validator deposits.
  • swETH Node Operator Registry (0x46DdC39E780088B1B146Aba8cBBe15DC321A1A1d): list of validators and operators.

The current on-chain implementation and admin of these three contracts were not read in the control data (see Technical Architecture). The rswETH sibling uses a near-identical design: it is "for the most part a direct fork of the swETH set of smart contracts", with EigenLayer added .

Yield Sources

Yield comes from Ethereum staking rewards earned by Swell's validators, not from token emissions. The swETH-to-ETH rate rises as rewards accrue. The docs' illustration is explicitly not the real rate .

Swell's docs split total rewards three ways. The staker keeps 90%. The node operator and the Swell DAO treasury get 5% each. That page is dated about two years before the snapshot. The rswETH contract on-chain still shows a 5% node operator share (nodeOperatorRewardPercentage: '50000000000000000', i.e. 5% in 18-decimal form) and a 5% treasury share . That confirms the split for rswETH only. The swETH contract's parameters were not read.

  • Sustainability: the base yield is protocol-level Ethereum staking yield, so it does not depend on emissions. A numeric historical APR range is not in the workspace (data gap).
  • Extra rewards: the restaking sibling also earns EigenLayer points and AVS rewards . That does not apply to swETH.
  • Exchange-rate caveat: a reported yield only counts if the rate reflects real ETH. In October 2026 press reports described the swETH rate as above its last clean level (see Key Risks).

Dependencies & Integrations

  • Ethereum validators and deposit contract: all staked ETH ends up in validators via the Ethereum deposit contract (see How It Works, step 3).
  • Node operators: a permissioned set run the validators. The DAO's planned role includes "Permissioned entry, onboarding, and relationship management of node operators."
  • Oracle for the exchange rate: the docs say the rate is "calculated in part by a Chainlink" proof-of-reserve oracle that reports the consensus-layer ETH balance daily . The on-chain rswETH contract also has a reprice(uint256,uint256,uint256) function. The control data does not resolve who can call it, and its delay is unknown, not zero . The contract sets a minimum interval of one hour and a maximum single change of 1% . The Defiant reports that the published swETH oracle code requires the BOT role to submit repricing snapshots (source reliability: unverified publisher). Assessment: the rate is likely moved by a privileged bot-role submitter within those caps, not only by a daily Chainlink feed. The basis is the reprice function and caps (measured for rswETH) and that press description of the swETH oracle code.
  • Cross-chain: DefiLlama shows TVL only on Ethereum. Swell's April 2026 post says it paused LayerZero bridging for swETH, rswETH, SWELL and rSWELL, and Nucleus bridging for earnETH, after the Kelp rsETH incident . On 2026-04-28 Swell said bridging was restored with verifier (DVN) settings raised from 1/1 to 3/3 . Both are X posts by Swell's account; the claims are not independently verified.
  • DeFi integrations: one third-party incident report (AI-written, not fully checked, source reliability: unverified publisher) says Aave, Spark and Fluid did not accept either token as of late 21 Oct, and a Morpho market priced off the contract rate had about 34 ETH to borrow . Unchained reports that a Uniswap swETH pool had roughly $143,000 of liquidity (source reliability: unverified publisher).

TVL & Adoption

DefiLlama shows $28.5M of TVL, all on Ethereum, in a snapshot collected on 2026-10-29.

  • Chain distribution: Ethereum only (100%) . No inactive chains are listed.
  • Trend: no dated historical TVL series is in the workspace. The 7-day change in the snapshot is about -1.8% .
  • Rate vs TVL: the October 2026 rates (see Key Risks) may inflate a TVL computed from the token rate. The sources do not say how DefiLlama prices the tokens, so the true ETH backing could differ from the headline.
  • Scale check (rswETH, a different product): its contract shows totalSupply: '10654080875035056121180' rswETH , about 10,654 tokens. Press reports put rswETH TVL near swETH's, about $29M each (source reliability: unverified publisher).

Technical Architecture

Control state is measured in full for rswETH and only partly for swETH. The code-measured control file covers rswETH contracts on Ethereum at block 26300146 (2026-10-29 18:18 UTC). It has no entry for the swETH token, its Deposit Manager or its Node Operator Registry. Separate direct reads of the swETH token are given below, and the swETH docs say rswETH remains the same as swETH apart from EigenLayer.

swETH token controls (direct reads, Ethereum):

  • Proxy admin: the swETH token (0xf951E335afb289353dc249e82926178EaC7DEd78) stores an EIP-1967 admin (the address allowed to upgrade a proxy) of 0x25eaf579ca2255faa5463c635eec28697b5b8846 at block 26300521 . Its implementation is 0xce95ba824ae9a4df9b303c0bbf4d605ba2affbfc at block 26300523 .
  • Admin owner: that proxy admin's owner() returns 0xba829003e6b5dd858fbed3945719939a6dc3073b at block 26300524 . This is the same address the control file shows as the Compound-style timelock with a 7-day delay on the rswETH side . Assessment: swETH upgrades likely also pass this 7-day timelock. The basis is the shared owner address; the swETH Deposit Manager and registry admins were not read.
  • Access control: AccessControlManager() on the swETH token returns 0x625087d72c762254a72cb22cc2eca40da6b95eac at block 26300525 . The holders of its pause and reprice roles were not read.

rswETH controls (Ethereum, block 26300146):

  • Upgrades: the rswETH token, DepositManager and AccessControlManager are proxies. A shared proxy admin sits behind a Compound-style timelock with a 604800-second (7-day) delay . The upgrade privilege resolves to that timelock .
  • Timelock proposer: a Safe multisig (a contract wallet needing several signers) that requires 4 of 7 signers .
  • Pausing: a pause contract and a 1-of-5 Safe can pause core, bot and withdrawal methods immediately . A single signer in that Safe can freeze the product. Unpausing needs the 4-of-7 Safe, also with no delay .
  • Repricing: the reprice function showed no resolved holder and an unknown delay . Who holds the reprice role, and how fast it can act, was not resolved from the measured data. Reports attribute the October 2026 incident to this path, so it is the key control to confirm.

Pause state now (Ethereum). The rswETH access-control contract showed coreMethodsPaused: true, withdrawalsPaused: true and bot methods paused at block 26300146 . A direct read of the swETH access-control contract at block 26300482 shows core methods paused and withdrawals paused at block 26300483 . The address read is the one the swETH token itself returns for AccessControlManager() (see above).

Exchange rates at those blocks. The control snapshot is dated 2026-10-29 18:18 UTC. The direct reads at blocks 26300482 and 26300483 were fetched on 2026-10-29 at about 19:27 UTC. The swETH getRate() returned 1202969727705006739 at block 26300482, about 1.203 ETH per swETH . The rswETH rate was rswETHToETHRate: '1137680064739836267', about 1.1377 . Press analysis put the last clean rates near 1.133 for swETH and 1.082 for rswETH (source reliability: unverified publisher). Unchained reported the same swETH rate on 23 Oct (source reliability: unverified publisher). Assessment: the NAV was likely still uncorrected at those 2026-10-29 reads. The basis is the unchanged swETH rate, the pause flags still set and no resume announcement in the sources. The clean-rate figures come only from an unverified publisher, and Swell has not published the correct value.

Governance. The DAO votes off-chain on Snapshot, and passed proposals are implemented by the Swell DAO . So token votes do not execute on-chain actions directly. The docs' "Phase 2" describes DAO control over contracts and parameters as a future stage .

Emergency mechanisms. Pause switches exist for core, bot, operator and withdrawal methods (rswETH data above), and Swell used them on 2026-10-23. Its post says:

rswETH fork: who upgrades and pauses (block 26300146)Open full view →
GOVERNANCETimelock, 7-day delayROLESafe, 4 of 7 signersROLEPauser Safe, 1 of 5CONTRACTAccessControlManager(proxy)CONTRACTrswETH token andDepositManageradministerscan upgrade, after7 dayscan upgrade, after7 dayscan pause at oncecan unpause atonce
  • (governance)
  • (role)
  • (role)
  • (contract)
  • (contract)
  • control
  • control
  • control
  • control
  • control

Key Risks (Protocol-Level)

Where funds sit. User ETH sits in validators via the Ethereum deposit contract, in Deposit Manager balances and in withdrawal buffers. Swell's October 2026 message says "All funds are safe" . That is a first-party statement; no independent confirmation is in the workspace.

  1. Active incident (highest priority). Press reports say Swell paused after reports of compromised repricing keys (source reliability: unverified publisher). Swell has not confirmed the cause or any link to a cloud-platform breach (source reliability: unverified publisher). The same article quotes Swell: "no ETH has left the swETH or rswETH withdrawal or deposit contracts" (source reliability: unverified publisher). About 2,513 ETH was called "not confirmed loss" by a security researcher (source reliability: unverified publisher). Impact on holders: they cannot redeem through Swell, and the DEX price of swETH was about 1.089 WETH against a contract rate of about 1.203 (source reliability: unverified publisher). Assessment: when Swell corrects the NAV, the swETH rate may fall from about 1.203, so holders should not assume 1.203 ETH per swETH on redemption. The basis is Swell's own statement that NAV must return "to the correct value" and the DEX price gap. Swell has not published a corrected rate or reopening date.
  2. Privileged key risk. The measured rswETH design lets a single signer in a 1-of-5 Safe pause the system with no delay. The reprice function has no resolved holder and an unknown delay in the control data. The rswETH contract caps one reprice at 1% and requires one hour between reprices . Press accounts say compromised keys raised the rate in about 1% hourly steps (source reliability: unverified publisher). Assessment: a misused reprice role could move the rate by up to 1% per hour, and the single-signer pause is the fast stop. The key-compromise account itself is unconfirmed, and the reprice-role holder is unknown.
  3. Exchange-rate and oracle dependency. swETH value depends on a reported consensus-layer ETH balance. If the reporting path is wrong or abused, the rate diverges from ETH backing, and any lender or DEX pool using the rate is affected.
  4. Audit coverage for the current swETH code is only partly shown. No audit in the supplied reports is shown to cover the current swETH token, Deposit Manager or registry:
    • SoftStack, 2022-03-11: covers contracts/swETH.sol, swDAO.sol, swNFTUpgrade.sol and Strategy.sol from Swell's early 2022 design. The audit text describes swETH "pegged 1:1 to ether", unlike today's rising rate . It has not been compared to the current deployed code.
    • SoftStack, 2022-05-08: covers a Balancer-based Vault, a different product.
    • Hexens, review dated 2024-02-12 (titled January 2024): swETH withdrawals and validator-exit logic. The code scope was not extracted from the report, so its match to deployed code is not confirmed.
    • Hexens, review dated 2024-03-12: the SWELL token Merkle drop and SWELL staking contracts . Not swETH.
    • Hexens, review dated 2024-04-03, and MixBytes, 2024-04-10: the L2 pre-deposit staking and zap contracts (MixBytes lists SimpleStakingERC20.sol and Zap.sol ). Not swETH.
    • The rswETH DepositManager and token are marked as having unconfirmed audit scope in the audit-coverage table.
    • Later audits found by web research (source reliability: unverified publisher for the Codespect pages): Codespect lists a "Swell Repricing Update" dated 2025-04-17 for Swell's repricing oracle and EigenLayer slashing on rswETH, with one Medium finding marked Fixed . A "Swell Withdrawal Update" dated 2025-04-14 is described as a minor update to Swell's liquid staking contracts for EigenLayer withdrawal integration, with no findings listed . A "Swell ETH Staking Deposit Bot" report dated 2026-07-07 covers an off-chain tool that deposits validators for swETH and rswETH .
    • Swell's docs audits page says swETH and rswETH audits were conducted by Sigma Prime and Codespect, and lists a Sigma Prime swETH report (v2.1) without a date . The Sigma Prime report text and the Codespect PDFs could not be read, so their scope and findings are not assessed. No audit of the access-control contract and no audit after the October 2026 incident was found.
  5. Withdrawal delays. swETH exits through the protocol typically take 1-7 days, with occasional delays of up to 12 days or more, according to Swell's docs (see How It Works, step 6). Large requests can exceed Swell's buffer and lengthen queues . Withdrawals are paused now. The rswETH withdrawal page gives waits of up to 28 days .
  6. Complexity. The footprint includes registry, deposit manager, access-control manager, bridges (LayerZero), the SWELL token and other family products. The control file lists 83 privileged-function rows for rswETH, and the parser could not tell who can call 58 of them .

Data Completeness

  • Available: DefiLlama snapshot (2026-10-29), Swell docs on staking, exchange rate, rewards split, swETH and rswETH withdrawals, rswETH design, address lists, six audit reports, three official X posts, a code-measured control file for rswETH, direct on-chain reads of the swETH token, and web research on the October 2026 incident, swETH withdrawals and later audits.
  • Missing for swETH: admin, owner and roles for the swETH Deposit Manager and registry; the holders of the swETH pause and reprice roles; the swETH timelock delay confirmed directly (it is inferred from the shared owner address); a historical yield and TVL series; the node operator list; any Blockworks (ex-Messari) report (the workspace has no articles).
  • Weak sources: incident detail comes from press (The Defiant, Unchained) and a third-party incident report, all with unverified publisher reliability. Swell's own X posts carry no cause. The October incident cause, size and any loss are unconfirmed.
  • Stale docs: the rewards split, exchange-rate and staking pages are marked as last updated one to two years before the snapshot. Their Chainlink description may not match how the rate is now set.
  • Later audits: the Sigma Prime reports and the Codespect PDFs could not be read. The Codespect findings above come from summary pages only.
  • Audit scope: four reports had no extractable code scope or an ambiguous or disagreeing scope review. They prove no coverage.
  • Source-code coverage: the deployed-code audit-coverage table lists no swETH components. rswETH components are marked with unconfirmed audit scope. Nothing here shows changed or unaudited code; the comparison was simply not made.

108 citations · click markers to view evidence

🔒

Code Security

6/10

Code Security Analysis: Swell Liquid Staking

Conclusion. For swETH (the liquid staking token), the workspace holds one review of the 2024 liquid staking update, by Hexens (1 medium and 7 low/informational findings, all fixed or acknowledged), plus the March 2022 SoftStack/Chainsulting protocol audit of an earlier swETH version, which has one acknowledged critical finding. The other four published reviews cover the Balancer vault, staking, merkle drop and L2 pre-deposit contracts; they are family context and not swETH coverage. The weak point is the link between the swETH reviews and the contracts running today: no audit could be tied to a deployed address, and source code is available for only 4 of 15 listed addresses, none of them swETH. The acknowledged critical finding from the 2022 protocol audit could not be checked against code. A second acknowledged finding, a medium in the separate vault audit, is family context and not a swETH finding. Confidence is moderate, because the current swETH contracts were not compared with any audited version.

Data Completeness

Findings ledger: 6 audits and 49 issue groups (finding groups are one issue each). Read in full: the findings ledger, the audit index, all six audit digests, the audit coverage map, the contract context, and the audit sections quoted below. Other audit text reached this report only through the ledger and digests.

Known gaps:

  • Source code was collected for only 4 of 15 listed addresses (Sourcify runtime matches); 11 addresses were not checked for source.
  • No audit names the deployed addresses, and deployed code was not compared with any audited version (the audits do not give a reliable contract-level scope that links to deployed addresses).
  • The digest of the March 2022 SoftStack protocol audit was built from the first 80,000 characters only, and the digest notes that the report text is truncated ().
  • Repository code for the two non-fixed findings was not located, so no code check is possible.
  • No bug bounty page could be opened; the only bounty evidence is a 2023 Swell blog post, whose current status is unverified.
  • The audit coverage map lists no collection gaps, but the inventory is incomplete: two audited contracts have no attributed current address (swNFTUpgrade.sol initialize and SwellBalancerVault).

Scope note: the Sourcify-matched contracts are the rswETH restaking contracts (DepositManager and RswETH). The protocol family lists swell-liquid-restaking, swell-earn and swell-l2-farm as separate members. The documentation page also lists earnETH and SWELL token addresses. Audit evidence is counted per product below and not transferred between products.

Audit Coverage

Six included audits, all trusted auditor-hosted copies (GitHub repositories of the auditors). The audit index tags each as family-wide (covering a component shared across the Swell family), but that tag alone does not show swETH coverage. Code scope is listed in the audit text for SoftStack protocol (2022-03) and MixBytes (2024-04), and extracted from section tables for SoftStack vault (2022-05, with an ambiguous extraction). The three Hexens reports give no contract-level scope list; the 2024-02 report names a repository commit (). Coverage by product, as far as digests and scope lists allow:

AuditDateSubjectRelevance to swETH (liquid staking)
SoftStack/Chainsulting, protocol2022-03-11Earlier major version of swETH
SoftStack/Chainsulting, vault2022-05-08Balancer vault, a separate product
Hexens, v3 liquid staking2024-02-12The audited update adds swEXIT withdrawals and node operator exit logic; the report does not show that this commit is the deployed version
Hexens, staking and merkle drop2024-03-12Not swETH
Hexens, L2 pre-deposit2024-04-03Not swETH (L2 pre-deposit)
MixBytes, staking2024-04-10Not swETH (L2 pre-deposit)

The Hexens January 2024 review is the only one whose digest names the current swETH, DepositManager and related contracts; it reports "1 medium and 7 low/informational findings" and states that all were fixed or acknowledged (). For swETH this is a single independent firm. The Hexens report names a repository commit but no contract-level scope and does not link that commit to deployed addresses. The swETH contract list therefore comes from the digest only and proves no coverage of a deployed address on its own. The report calls the audited code "the next update of Swell Network" (), so whether it is the current major version is an inference from the contract names. If it is not, the swETH review would be of a previous version only; the audit-depth level is the same either way.

Unaudited or unconfirmed areas:

  • No included audit names the RswETH (rswETH token) contract; the rswETH deposit manager is not shown in any audit scope either.
  • The earnETH vault and SWELL token have no audit in this workspace. earnETH belongs to another family product (swell-earn) and is context only.
  • The EigenLayer-related restaking contracts have no audit in this workspace. The Hexens 2024-02 digest references an AccessControlManager.sol (), but no source shows that this is the access control manager now deployed for swETH or rswETH, so its audit coverage is unconfirmed.
  • Family-level audits of other Swell products (swell-liquid-restaking has 1 audit and 1 doc recorded) are not counted here.

Deployed code vs audits

Deployed code was not compared with any audited version: 100.0% of 4,770 physical lines in 22 protocol-owned verified files are "unknown" (0 identical, 0 cosmetic, 0 changed, 0 unaudited), because the audit scopes could not be tied to the deployed contracts (). Lines in changed, unknown or unaudited code, by reviewed criticality, are: critical 1,392, high 746, peripheral 172, unrated 2,460. The assessed roles are funds, access control, upgradeability, oracle and view utilities. Some code has not been assessed for criticality, and 11 components were not measured. No files lack line counts.

Sourcify matches the runtime bytecode of the DepositManager implementation (0xee33…3181, behind proxy 0x5e63…4889) and the RswETH implementation (0x4796…2362, behind proxy 0xfae1…c0) to the listed source files (). This says nothing about audit scope or commits.

ComponentDeploymentStatusAudit and revisionCriticalityNote
DepositManager (rswETH deposit manager v1.0)eip155:1 impl 0xee33…3181, proxy 0x5e63…4889Audit scope not confirmed; version not comparedNone names itCritical (code rating)Sourcify runtime match; the audit scope could not be determined from the available material
RswETH (rswETH token v1.0)eip155:1 impl 0x4796…2362, proxy 0xfae1…c0Audit scope not confirmed; version not comparedNone names itHigh (code rating)Sourcify runtime match; same scope gap
swETHeip155:1 0xf951…ed78Source not collectedHexens 2024-02 digest names swETH.sol; no link to the addressCriticalVersion not compared
swETH Deposit Managereip155:1 0xb3d9…8f39Source not collectedHexens 2024-02 digest names DepositManager.solCriticalVersion not compared
swETH Node Operator Registryeip155:1 0x46dd…1a1dSource not collectedHexens 2024-02 digest names NodeOperatorRegistry.solHighVersion not compared
rswETH Node Operator Registryeip155:1 0xaae0…c524Source not collectedNoneHighVersion not compared
EigenPod (rswETH)eip155:1 0x8d0b…e7258Source not collectedNoneCriticalHolds restaked validator ETH
earnETH vaulteip155:1 0x9ed1…5f22Source not collectedNoneCriticalAnother family product
earnETH managereip155:1 0x69fc…50c0Source not collectedNoneHighAnother family product
earnETH accountanteip155:1 0x411c…2755Source not collectedNoneHighAnother family product
earnETH tellereip155:1 0x685a…0429Source not collectedNoneCriticalAnother family product
earnETH rolesAuthorityeip155:1 0x1f5d…a016Source not collectedNoneHighAnother family product
SWELL tokeneip155:1 0x0a6e…5676Source not collectedNonePeripheralGovernance token
swETH (audit-named)No attributed addressNo address to compareSoftStack 2022-03CriticalEarlier version
swNFTUpgradeNo attributed addressNo address to compareSoftStack 2022-03HighInventory gap
SwellBalancerVaultNo attributed addressNo address to compareSoftStack 2022-05HighInventory gap
SimpleStakingERC20No attributed addressNo address to compareHexens 2024-04; MixBytes 2024-04CriticalL2 pre-deposit
ZapNo attributed addressNo address to compareHexens 2024-04; MixBytes 2024-04HighL2 pre-deposit
StrategyNo attributed addressNo address to compareSoftStack 2022-03HighEarlier version
swDAONo attributed addressNo address to compareSoftStack 2022-03PeripheralEarlier version
helpers, ISWNFT, NFTDescriptor, NFTSVG, HexStringsNo attributed addressNo address to compareSoftStack 2022-03Peripheral5 helper and library names
ERC20, ERC4626, FixedPointMathLib, SafeTransferLib, IVault, IWETHNo attributed addressNo address to compareSoftStack 2022-05Peripheral6 vendored library names
ghtedPoolFactory, rVault, ol, h, s, y, ptorNo attributed addressNo address to compareSoftStack 2022Peripheral7 truncated file-name fragments from the audit tables

Count: 13 deployed components and 25 inventory names, 38 rows in total; the table shows them grouped, so no rows are omitted. The address list in the protocol documentation is first-party, but the inventory is partial: the two audited contracts swNFTUpgrade.sol (initialize) and SwellBalancerVault have no attributed current address ().

Contract controls observed on-chain

On Ethereum at block 26300146, the upgrade right over the rswETH DepositManager proxy sits with a timelock of 604,800 seconds (7 days) (). The same state shows coreMethodsPaused, botMethodsPaused and withdrawalsPaused all true on the access control manager (). Separately, Swell announced on 2026-10-23 that deposits and withdrawals for swETH and rswETH were paused (). This event post-dates every audit in the workspace.

Findings by Severity

Source note: the press report on the October 2026 pause comes from an unverified publisher; Swell's own posts and the audits carry the factual claims here.

Critical

  • Initialize not protected (SoftStack/Chainsulting, 2022-03-11, finding 5.1.1; swNFTUpgrade.sol) — — Audit status: acknowledged. The team's update says the function is guarded at proxy level and storage is not read on the implementation () — Check: not checked: the described code was not found in the collected repository code.
  • Critical issue reported in the executive summary, not itemized (SoftStack/Chainsulting, 2022-03-11) — Audit status: fixed.

High

  • Possible reentrancy attack vector (SoftStack/Chainsulting, 2022-03-11) — Audit status: fixed. High issue in the executive summary, not itemized — fixed.

Medium

  • Rounding errors can lead to ignore slippage (SoftStack/Chainsulting vault audit, 2022-05-08; SwellBalancerVault.sol; separate Balancer vault product, family context and not a swETH finding) — — Audit status: acknowledged. The update says withdrawals below 100 revert because of a Balancer minimum-exit condition and a unit test was added () — Check: not checked: the audit's repository code was not available.
  • Fixed medium groups (3): missing require check and an executive-summary medium (SoftStack/Chainsulting, 2022-03), and ERC721 safe mint reentrancy in withdrawals (Hexens, 2024-02, ).

Low / Informational

The ledger holds 49 groups in total. Besides the 2 critical, 2 high and 4 medium groups above, 41 are low (27) or informational (14). Most are fixed. Acknowledged ones are:

  • Error strings in require, hardcoded address (SoftStack, 2022-03)
  • For loop over dynamic array, missing NatSpec documentation (SoftStack, 2022-05)
  • Withdrawal pause without time limitation (Hexens, 2024-02): the digest says pauseWithdrawals has no automatic resume, creating a centralization risk of indefinite withdrawal freezing (). The October 2026 pause of deposits and withdrawals makes this relevant to users; the April 2026 pause was of bridging only ().

One informational item has no stated status, and its severity is estimated: the claim that the deployer/owner cannot pause the contract was marked failed (SoftStack, 2022-03). The rating is an AI estimate, because the audit states no severity; the failed claim could mean the owner can pause, a centralization point with limited direct fund impact.

Recurring Patterns

  • ERC721 _safeMint reentrancy appeared in two audits as separate issues: a high finding in swNFTUpgrade (2022, fixed) and a medium finding in swEXIT (2024, fixed). They are different code and different groups, but the same bug class.
  • Pause and owner-control power: the 2022 failed claim about owner pausing and the 2024 unlimited withdrawal pause are separate groups on the same theme.
  • Missing zero-address checks appear in both 2022 SoftStack audits (fixed).

Code Quality Assessment

Based on available data: for swETH, the 2024 Hexens review reports 1 medium and 7 low/informational findings with no critical or high, all fixed or acknowledged. The earlier-version 2022 protocol audit left one critical issue acknowledged; the audit-level fix rate of the other reviews (vault, staking, L2 pre-deposit) is not counted here. The one critical issue (2022) was acknowledged with a stated proxy-level mitigation, which was not code-checked. Weaknesses are structural: the audited contracts are not tied to deployed addresses, the most recent audit of swETH is from February 2024 while the contracts have been exposed to an operational incident in October 2026, and the rswETH contracts (the only ones with collected source) are not named in an audit scope. A bug bounty existed in 2023 with a reported critical maximum of USD 100,000, capped at 10% of funds at risk; its current status could not be confirmed () (Swell's own 2023 blog post).

One report on the October 2026 pause attributes it to compromised repricing keys, but the available Swell sources do not confirm a cause () (source reliability: unverified publisher). The cause stays unconfirmed. If compromised keys were the cause, it would be an operational key-control matter rather than a finding about audited code; it is not scored as a code finding here.

Aggregators: no CertiK, DeFiSafety or Exponential assessment is present in the workspace.

Score: 6.0/10

Rubric security-score-v2.6; data completeness 0.85

Code security factors, points out of 10 (points)
Audit coverage
5.50 points
Audit depth
5.50 points
Unresolved findings
8.68 points
Assurance
3.50 points
Audit findings not marked fixed (findings)
Critical · acknowledged
1 findings
Medium · acknowledged
1 findings
FactorLevelEvidenceWeightPoints
Audit coverage (computed)L2supported0.355.50
Audit depthL2supported0.205.50
Unresolved findings (computed)L4supported0.308.68
Assurance—not disclosed0.153.50

26 citations · click markers to view evidence

💰

Economic Security

3.9/10

Economic Security Analysis: Swell Liquid Staking

Conclusion. Swell's liquid staking (swETH) and liquid restaking (rswETH) tokens are in an unresolved incident at the research date (2026-10-29). Withdrawals are paused on-chain for rswETH and swETH, and Swell says the next step is to return NAV (net asset value, the exchange rate) to the correct value. Treat the position as illiquid at the protocol level until Swell resumes exits. A targeted search found no Swell incident report or post-mortem, so the cause stays unconfirmed. Press and analyst accounts of a key compromise and of inflated rates are unverified. Whether holders lose principal depends on facts that are not yet public.

TVL & Liquidity Profile

The protocol is small and sits entirely on Ethereum, and the only measured DEX depth, for rswETH, is zero.

  • The latest captured DeFiLlama snapshot (collected 2026-10-29) shows TVL of about USD 28.5M, all on Ethereum . Ethereum is the base chain and has no L2 exit window .
  • The same-day composition is 100% WETH . Only rswETH is measured. Its market cap is about USD 28.3M . The mapping of the DeFiLlama record to swETH, rswETH or both is not stated. A press report (source reliability: unverified publisher) says DefiLlama lists swETH at about USD 29M and rswETH at roughly the same . Total exposure across both tokens may therefore be larger than this one record.
  • DEX depth for rswETH at 2% price impact is USD 0 with no venue found. A USD 100,000 probe moved the price 19.41% . The 0.5% depth (depth_usd_50bps) is null, which is a gap, and as a coarse-probe lower bound it adds nothing. swETH depth is not measured.
  • A press report (source reliability: unverified publisher) cites a swETH Uniswap pool with roughly USD 143,000 of liquidity in an October 23 snapshot; current swETH depth is unmeasured .
  • Native exit queues are long. rswETH withdrawals take up to 28 days . swETH withdrawals typically take 1-7 days, with delays of 12 days or more .
  • Top-holder concentration for swETH and rswETH is not in the workspace. Insufficient data for this assessment.

Oracle Risk

The exchange rate is pushed by a bot key under an on-chain 1% step limit. An unverified press and analyst account alleges a key compromise that stayed within that limit by repeated steps.

  • Design. The swETH rate is computed in part from a Chainlink proof-of-reserve oracle for consensus-layer ETH, updated daily . The rate drives swETH minting and ETH returned on unstake. The oracle contract's bot (BOT role) submits the reserve figure for repricing .
  • On-chain bounds, rswETH, at Ethereum block 26300146 (2026-10-29). The reprice step limit is 1% . The minimum time between reprices is 3,600 seconds . The contract reports 1.1377 ETH per rswETH . The last reprice was at Unix time 1792643135, which is 2026-10-22 04:25 UTC .
  • swETH rate. getRate() returned 1.2030 ETH per swETH at block 26300482 .
  • Audit history. Hexens (Jan 2024) found that the oracle used the proof-of-reserve answer without a staleness check. The audit marks that finding fixed . A Codespect repricing-update audit of 2025-04-17 found a Medium double-counting of slashing, marked Fixed (source reliability: unverified publisher).
  • Incident (weak sources). Press reports (source reliability: unverified publisher) cite an on-chain analyst saying admin keys looked compromised and NAV was raised "max 1% (unbacked) each hour" . The same source gives the contract rates it read (about 1.203 ETH for swETH and 1.138 ETH for rswETH) and says they sit roughly 6% and 5% above the last rates the analyst Spreek identified as clean . The clean baseline, the compromise and the cause come from that account, not from Swell or a non-weak source, and stay unconfirmed. Only the current contract rates are measured on-chain. A third-party GitHub report by Spreek (source reliability: unverified publisher) also states that an unknown actor raised both rates by 1% an hour . It is an inference by its author, not a Swell statement. A second source says the bot pause on both tokens is active, but that no role-revocation event has appeared (source reliability: unverified publisher). Swell's own post says only that returning "NAV to the correct value" is the next step .
  • Control of the updater key. The on-chain read lists role holders as a contract (0xd5a7…) and Safe multisigs. One of them has threshold 1 of 5 owners . The role hashes are not labelled, so the holder of the repricing (BOT) role is not identified by the workspace.
  • Per-product bounds. The 1% step and 3,600-second minimum are read on-chain for rswETH only. No source in the workspace gives a current on-chain bound for the swETH rate. The Hexens audit (Jan 2024) of the swETH system describes a bot-reported reserve checked against the proof-of-reserve oracle, which is the audited version and not a current bound. Who holds the swETH repricing key and whether a current on-chain bound applies to swETH remain open gaps; neither is inferred here.
  • Assessment: the rswETH rate is pushed by a bot role, bounded on-chain by a 1% step and a 1-hour minimum. These bounds cap the speed of any rate increase but not its total, since repeated steps compound. Without a first-party post-mortem, the compromise claim stays unconfirmed.
  • There is no spot-DEX price input in the minting and redemption design. Fallback if the oracle fails: the only documented mechanism is a pause (see Solvency). The sources state no time limit or reopening date for the pause.

Flash Loan & MEV Exposure

A flash-loan round trip on mint and redeem is blocked by queued withdrawals. The real exposure is a mispriced rate, not same-block manipulation.

  • Withdrawals go through a queue: 1-7 days for swETH and up to 28 days for rswETH (sources above). That makes a single-transaction deposit-and-redeem loop impossible by design. Assessment: flash-loan manipulation of mint/redeem is unlikely, based on the queue plus the hourly reprice limit.
  • If a contract rate sits above the true NAV, redeeming at that rate would pull more ETH per token than the pool backs, and depositing would mint fewer tokens. Swell paused both paths, which is consistent with this; whether the rate is above the true NAV rests on the unverified account above. A report on swETH (source reliability: unverified publisher) says its withdrawal contract pays the lower of the request-time and processing-time rate, and the current scope for rswETH is not confirmed .
  • Audit evidence: Hexens (Jan 2024) found a medium reentrancy in swEXIT.createWithdrawRequest, where a BOT-role holder could drain the withdrawal queue. The audit marks it fixed . The same audit's low-severity finding says pauseWithdrawals has no automatic resume and was acknowledged .
  • Sandwich and MEV exposure on the secondary market is likely for rswETH because its measured DEX depth is zero (see above); swETH depth is unmeasured. Swell's docs warn that the secondary-market rate may differ from the primary rate and may carry slippage .
  • Liquidation front-running does not apply to the base mint/redeem path. Downstream lending use is a separate exposure (see Economic Attack Vectors).

Tokenomics Risk: Yield, Leverage & Reflexivity

User return is Ethereum staking rewards net of a 10% fee; there is no leverage in the user position and no own token in the backing, but the revenue-versus-emission split is not quantified.

  • Named revenue. For swETH the docs split total rewards 90% to stakers, 5% to node operators and 5% to the Swell DAO treasury . The rswETH contract carries matching 5% and 5% parameters on-chain . The workspace has no cited ETH amount of rewards paid, no APY figure and no fee total.
  • Emissions and points. rswETH holders also get off-chain EigenLayer points . SWELL token airdrops went to early stakers and liquidity providers . The share of paid yield from own-token emissions is not cited, so the revenue/emission split is a gap.
  • Leverage and looping. The mint and redeem path has no leverage. Users who loop swETH or rswETH in other lending protocols hold that leverage outside Swell, and the workspace has no LTV data for it.
  • Own token in backing. Composition is 100% WETH , so the composition read shows no SWELL in the pool. SWELL has a maximum supply of 10,000,000,000 . The team allocation vests over 36 months . SWELL unlocks do not touch the backing and do not score here.
  • Reflexive path. The loop at issue is the rate itself. If NAV were set above the backing, redemption value would rise and a rate attacker could exit at unbacked value; the sources do not confirm that this happened. This is a rate-integrity risk rather than an own-token spiral.

Solvency, Liquidation & Bad Debt

Exits are halted at the research date, so principal is currently inaccessible through the protocol and its solvency cannot be shown.

  • Halt confirmed on-chain. For rswETH at Ethereum block 26300146 (2026-10-29 18:18 UTC), bot methods, core methods and withdrawals are all paused . For swETH, the access-control contract reports withdrawalsPaused() true at block 26300483 . Swell's post announces the pause for both tokens .
  • No stated bound. Swell gives no reopening date. The Hexens audit of an earlier version recorded that the contract then lacked automatic resumption of withdrawals, and marked the finding acknowledged; whether the current contract differs was not compared . Swell gives no reopening date; a press report (source reliability: unverified publisher) says the same .
  • Backing. The pool holds ETH on the consensus layer and in EigenLayer pods. The rswETH contract reports 147,044 ETH deposited in total and totalSupply of about 10,654 rswETH . The first figure is cumulative deposits, so it cannot be compared with current backing. A backing-versus-liabilities ratio is not measured. No ERC-4626 vault read exists.
  • Loss claims. Swell says "All funds are safe", and an early statement says no ETH left the deposit or withdrawal contracts (source reliability: unverified publisher). An analyst says the 2,513 ETH reported held is "NOT confirmed loss" (same source). No bad debt or uncovered shortfall is cited. Absence of a cited shortfall is not proof of full backing.
  • Debt positions. Swell itself runs no lending or liquidation. Downstream use is a separate exposure. A press report (source reliability: unverified publisher) says Aave, Spark and Fluid did not accept either token as of late Oct. 21 .
  • Withdrawal delay when running. Normal exit is up to 28 days for rswETH . Swell says large requests can exceed its buffer of rewards and daily deposits .

Incentive Alignment

Swell's operators earn 5% and the DAO treasury earns 5% of rewards, so the team is paid by volume. Users carry the full tail risk of key compromise.

  • Loss cover. Docs describe service fees as directed to "insurance and development funds" that the DAO distributes . This is a roadmap statement. The workspace gives no fund size, no address, no payout trigger and no pause rule, so it is not recorded as a loss-cover fund.
  • Who paid past incidents. The DefiLlama hacks list for this record is empty . The 2026-10 incident is not in that list and has no payout record. The April 2026 LayerZero/Kelp event led to a bridging pause; Swell then raised its verifier setup from 1/1 to 3/3 . No loss at Swell is cited from that event.
  • Mercenary capital. TVL of about USD 28.5M with zero measured rswETH DEX depth suggests that sales of rswETH on DEXs are exposed to heavy slippage: a USD 100,000 probe moved the price 19.41%. The probe measures price impact, not the market price against redemption value. The share of TVL tied to a points campaign is not cited.
  • Death-spiral risk. Not an own-token spiral. A confidence run is possible if the corrected rate ends up below the current contract rate: holders would race to exit at whichever rate is honoured. A report on swETH (source reliability: unverified publisher) describes a pay-the-lower-of-two-rates rule that would limit gains from a high request-time rate; its current scope and effect are unconfirmed.

Economic Attack Vectors

The rate-update key is the main vector the sources point to; secondary vectors are staking-layer dependencies and thin exit liquidity.

  • Price/rate manipulation. A bot key can raise the rate 1% per hour. A press report cites four malicious upgrade contracts able to take about 2,500 ETH that had not been queued through the multisig (source reliability: unverified publisher). Upgrades on the rswETH core contracts pass a 7-day timelock (min delay 604,800 s) . The EigenPodManager upgrade path (an EigenLayer contract, not Swell's) is controlled by a 1-of-2 Safe with no delay .
  • Liquidation cascades. Swell holds no debt. Any lending market that prices swETH or rswETH off the contract rate would value collateral too high if the rate sits above the true NAV. A press report says a Morpho market priced off the contract rate had about 34 ETH available to borrow (source reliability: unverified publisher). Impact is limited by the small size cited.
  • Bad debt. None is cited for Swell. Downstream bad debt would sit with those protocols.
  • Bridges. swETH, rswETH, SWELL and rSWELL use LayerZero bridging, and earnETH uses Nucleus bridging; Swell paused both after the April 2026 Kelp incident . All current TVL is on Ethereum, so bridge exposure is for bridged holders only.
  • Composability. Every downstream protocol that accepts either token inherits the rate risk. For rswETH the measured DEX depth is zero, so a large rswETH sale would meet heavy slippage (a USD 100,000 probe moved the price 19.41%); swETH depth is not measured and the cited USD 143,000 pool figure is an unverified snapshot of 2026-10-23.

Dependency risk (critical dependencies)

  • EigenLayer (rswETH): withdrawals depend on its 14-day delay and slashing . rswETH uses EigenLayer's EigenPodManager, whose admin is a different Safe (see above).
  • Ethereum consensus layer: exit queue and validator sweep delays.
  • Chainlink proof-of-reserve oracle (swETH rate input).
  • Permissioned node operators: operator onboarding is permissioned and overseen by the DAO .
  • Off-chain bot and key infrastructure: the repricing key, whose compromise is alleged by press reports (source reliability: unverified publisher).
  • LayerZero and Nucleus bridging: the April 2026 post shows DVN configuration was 1/1 before the upgrade to 3/3.
  • aggregators/dependency_graph.yaml lists no scored protocol dependency, only unresolved WETH, and warnings is empty .

Aggregator Scores

No curated security score is available. The workspace has no CertiK, Exponential or DeFiSafety file, so there is nothing to compare with the detailed analysis. The only aggregator inputs are the DeFiLlama TVL snapshot (collected 2026-10-29), DEX depth and token reads, which are described above. The DefiLlama hacks list is empty and does not reflect the October 2026 incident.

Data Completeness

  • Available: TVL snapshot; on-chain pause and parameter reads for rswETH (block 26300146); web-fetched on-chain reads for swETH (blocks 26300482-26300525); mechanism docs; audits of earlier Swell versions; Swell's own X posts about the incident.
  • Missing: a first-party incident report or post-mortem (a web search on 2026-10-29 found none: the Swell blog and homepage captures show no October 2026 incident post, and no official update after 2026-10-23 or reopening date was found); ETH amounts of yield paid and APY; emissions share of yield; top-holder concentration; DEX depth for swETH; backing-versus-liabilities measurement; insurance fund size and address; identity of the repricing key holder; any current bug bounty status (a 2023 Immunefi post says rewards of up to USD 100,000 , and its current status was not checked).
  • Source quality: the incident details come mainly from press reports (source reliability: unverified publisher) plus Swell's own post and on-chain reads. The pause state is established on-chain; the key-compromise cause and the amount of any loss are not.
  • Confidence: high that exits are halted at the research date; moderate for the oracle and dependency calls; low for yield design and for liquidity concentration beyond DEX depth.

Score: 3.9/10

Rubric security-score-v2.6; data completeness 1.00

Economic security factors, points out of 10 (points)
Solvency redemption
1.00 points
Oracle design
5.50 points
Dependency risk
5.50 points
Liquidity concentration
5.50 points
Yield design
5.50 points
FactorLevelEvidenceWeightPoints
Solvency redemptionL0supported0.351.00
Oracle designL2supported0.255.50
Dependency riskL2supported0.155.50
Liquidity concentrationL2supported0.105.50
Yield designL2supported0.155.50

58 citations · click markers to view evidence

👥

Team & Transparency

5.7/10

Social & Team Transparency: Swell Liquid Staking

Swell is run by an identifiable legal entity, publishes its contract addresses and audit reports, and has been live since at least April 2023 . The DefiLlama hacks dataset holds no record for the protocol or its family . An October 2026 event now dominates the picture: a withdrawal pause on swETH was read on-chain on 29 October 2026. Swell's own X posts (source reliability: unverified publisher) attribute it to a compromised operational key that inflated exchange rates, but no source of higher reliability confirms that cause, a root cause, a reopening date or a loss figure. Users should treat swETH exits as halted until Swell reopens them.

Team Identity

Swell has a named legal operator with a stated jurisdiction, but only one named leader, and that leader's profile rests on third-party pages of lower reliability.

  • Legal entity. Swell's terms of service say the protocol is "operated by or on behalf of LD Technologies Foundation, a company incorporated in the Republic of Panama" . The terms are dated 18 October 2024, so current control of the entity was not re-checked.
  • Other entities. The same terms name "Overflow Foundation" in the SWELL token section . The available sources do not explain how Overflow Foundation, LD Technologies Foundation and Swell Labs relate. The docs also refer to compliance with "the Swell Foundation" .
  • Named leader. Daniel Dizon is described as CEO and co-founder of Swell (source reliability: unverified publisher; unconfirmed by a first-party source) . A RedStone case study (source reliability: community publication) calls him "Swell Labs CEO & Founder" .
  • Background (unconfirmed). The same wiki (source reliability: unverified publisher) says he worked as a solidity engineer at Synthetix and dHedge before co-founding Swell . Brave New Coin (source reliability: unverified publisher) adds that he "founded and exited a software start-up" and lists backers including Mark Cuban, Bankless, Framework, Apollo and Maven 11 .
  • Gaps. Swell's own team or about page could not be retrieved. Other staff names seen only in search snippets were not confirmed. Only one leader is named, and that name comes from sources that are not first-party.
  • Token allocation. Docs allocate 25% of SWELL supply to the team, vesting over 36 months after a 12-month cliff . A Nansen page says the team is based in Sydney, with no founder names (third-party, not first-party).

Track Record

The family has been listed since 27 April 2023. The record now includes an October 2026 swETH withdrawal pause of unconfirmed cause and a reported precautionary bridge pause in April 2026.

  • Operating history. The family was first listed on 27 April 2023. The contract-creation measurement covers only the rswETH v1.0 contracts of the separate liquid restaking family member . The creation date of the swETH contracts is not measured. Code derives the exact family and version ages.
  • Core contracts. First-party docs list the swETH token and the swETH deposit manager . The same docs list rswETH contracts, which belong to the separate liquid restaking family member and are not scored here.
  • Launch security process. Swell's 2023 bug-bounty post says the launch followed "a testnet launch, guarded private launch, and a smart contract audit from leading provider Sigma Prime" .
  • Roadmap delivery. The March 2024 Barracuda upgrade enabled swETH withdrawals at the primary-market rate . In April 2026 Swell announced the sunset of its own L2, Swellchain, to focus on Faro . The post states that swETH, rswETH and SWELL on Ethereum are not affected. Swellchain itself is not in this protocol's assessed record.

Reported incident (unconfirmed): swETH and rswETH exchange-rate key compromise, October 2026

  • What is reported. Swell's pinned X article (source reliability: unverified publisher) says "an attacker used a compromised operational key to inflate the swETH and rswETH exchange rates by about 6% and 5%". This account is unconfirmed: no source of higher reliability states the cause . Only the headline of that article was captured. Because the account rests on unverified publishers, it is not recorded as an incident row; the on-chain pause below is the confirmed part.
  • Swell's statements. On 22 October Swell said "No ETH has left the swETH or rswETH withdrawal or deposit contracts" (source reliability: unverified publisher) . On 23 October it said it had paused deposits and withdrawals and that "All funds are safe" (source reliability: unverified publisher) . NAV (net asset value) is the ETH value each token represents.
  • On-chain state. At block 26300483 on Ethereum, the swETH access-control contract (0x625087d72c762254a72cb22cc2eca40da6b95eac) reported withdrawals paused . At block 26300482 the swETH contract's getRate() returned 1.2029 ETH per swETH . For the separate liquid restaking product (rswETH, another family member), its access manager (0x796592b2092f7e150c48643da19dd2f28be3333f) reported withdrawals, core methods and bot methods paused at block 26300146 on 29 October 2026 .
  • Researcher and press claims. Unchained (source reliability: unverified publisher) reports the rates were roughly 6% and 5% above the last clean values, and that Swell "has given no reopening date or confirmed a cause" . Researchers raised a possible link to a Railway cloud-platform breach, which Swell has not confirmed . The reported 2,513 ETH figure is "NOT confirmed loss" per the same source.
  • Assessment: The on-chain pause confirms that swETH withdrawals were halted at block 26300483, and the pause was posted as starting on 23 October. The key-compromise cause, the rate inflation and the absence of lost ETH come from Swell's X posts and press reports (unverified publishers) and stay unconfirmed. Whether holders face a loss depends on the NAV correction, and no source gives its terms.
  • Remediation. No reopening, role revocation or post-mortem is confirmed in the sources. A Defiant read of 23 October found no role-revocation event (source reliability: unverified publisher) .
  • DefiLlama hacks. The hacks dataset has no record for this protocol or its family . The October incident is therefore not yet in that dataset.

Earlier stress: Kelp rsETH incident, April 2026

  • On 19 April 2026 Swell said it paused LayerZero bridging for swETH, rswETH, SWELL and rSWELL, and Nucleus bridging for earnETH, as a precaution (source reliability: unverified publisher) .
  • Swell's 28 April X post (source reliability: unverified publisher) says it raised its verifier (DVN) configuration "from 1/1 to 3/3" . The post is the only source for the earlier 1/1 setting, so that configuration and the date bridging resumed are unconfirmed. No loss is reported.

Own tokens and depeg history

  • The only token in the asset checks is rswETH, which belongs to the separate liquid restaking family member. It is family context and is not listed as this protocol's own token. Its depeg history is not applicable because its exchange rate was unavailable to the check .
  • Press reports (source reliability: unverified publisher) say swETH traded near 1.089 WETH in a pool with about USD 143,000 of liquidity, against an on-chain contract rate of about 1.203 . This is a gap between the pool price and the on-chain contract rate, not a measured depeg against par. The press accounts that the contract rate was inflated stay unconfirmed.
  • swETH is not in the measured asset set, so its depeg history is not measured.

Social Presence

Development and communication continue, but at a low and irregular cadence. No community size data was available for Discord or Telegram.

  • GitHub. The public v3-core-public repository's master branch lists commits on 22 October 2026, 8 July 2026, and before those 13 November 2024 . That is commits in 2 of the last 12 months, and 19 commits in total on the branch. Contributor counts did not render. Visible authors are tiendv89, swellnetworkio and aaron-alderman.
  • Unexplained commit. The 22 October 2026 commit is titled "ci: zz-w22-s08 secret dump (org-secret proof)" . The GitHub capture displays the committing account as tiendv89, but the sources do not identify the account holder or say why the commit was made. It falls in the window of the reported October event, and the sources do not connect it to that event.
  • X (Twitter). A logged-out capture shows 660 followers, 60 posts and "Joined May 2025" (source reliability: unverified publisher) . The join date suggests the account may have been re-created, but the sources do not say so. Dated posts appear on 5, 16, 22 and 23 October 2026 and on 19 and 28 April 2026.
  • Blog. The cached blog shows 2026 posts on 6 March (fee-flow auctions) and 28 April (Swellchain sunset) . No later blog post was captured, and the cache date is unknown.
  • Discord and Telegram. The X bio links a Telegram invite. No member counts or discussion quality data were collected. Swell's official blog links a Discord invite .

Transparency Indicators

Addresses and audits are public, but no periodic treasury reports were found, the bug-bounty status is unconfirmed, and the overall open-source status is not established.

  • Addresses. Official docs publish contract tables for swETH and rswETH, including node operator registries . The deployed-addresses page also lists the SWELL token and earnETH contracts.
  • Audits. Docs state that "All audits are available in this repository" and name Sigma Prime and Codespect as auditors of swETH and rswETH . Swell also states Barracuda was audited by "Sigma Prime, Cyfrin, and Hexens" . A Codespect report dated 7 July 2026 on the ETH staking deposit bot lists one high finding, marked fixed (source reliability: unverified publisher) .
  • Bug bounty. Swell launched an Immunefi program in April 2023 paying "Up to USD 100 000" for critical issues . The program's current status could not be confirmed. A 29 October 2026 capture of the Immunefi listing shows no Swell row.
  • Financials. Docs describe allocations, vesting and the fee flow, but no periodic treasury, revenue or expense reports were found in the available sources.
  • Open source. The v3-core-public repository is public. The available sources do not show how much other organization code is public or private, so the overall open-source status is not established.
  • Incident communication. Swell's X account (source reliability: unverified publisher) carries an incident update. Only the headline of that update was captured, so its depth and any root-cause finding could not be checked.

Blockworks (ex-Messari) Article Freshness

No Blockworks (ex-Messari) reports are in the workspace, and the article index is empty.

Red Flags

  • swETH withdrawals were paused on-chain on 29 October 2026. Swell's own posts (source reliability: unverified publisher) attribute the event to inflated exchange rates from a compromised operational key, which is unconfirmed.
  • No confirmed root cause, no reopening date and no key revocation are visible in the sources.
  • Only one named leader is identified, and the identification comes from non-first-party pages. Swell's own team page could not be retrieved.
  • The relationship between LD Technologies Foundation, Overflow Foundation, Swell Labs and any DAO is not documented in the available sources.
  • Public commit cadence is low, and the 22 October commit title above is unexplained.
  • A Swell X post (source reliability: unverified publisher) says the bridge ran with a single verifier until April 2026; this is unconfirmed.
  • The bug-bounty program's current status is unconfirmed.

Positive Signals

  • A named legal entity with jurisdiction appears in the official terms of service.
  • Live since at least April 2023, with no record in the DefiLlama hacks dataset.
  • First-party address tables for swETH, and docs that name Sigma Prime and Codespect as auditors of swETH and rswETH .
  • Swell stated on X (source reliability: unverified publisher) that no ETH had left the contracts, with a stated plan to restore NAV before reopening. These statements are unconfirmed.
  • Swell stated on X (source reliability: unverified publisher) that it moved bridge verifier settings to 3/3 after the April 2026 Kelp incident. This is unconfirmed.

Data Completeness

  • Available: official docs and contract tables, a legal-entity terms page, GitHub commit history for one repository, official blog posts, X posts, on-chain reads at block 26300146–26300525, audit reports and several press pieces.
  • Missing or unconfirmed: Swell's team page, other named leaders, DAO or foundation charters, Discord and Telegram data, reliable follower history, the full text of the October incident update, a post-mortem, current bug-bounty terms and periodic financial reports.
  • The family context and previous-run files were not present. This analysis rests on collected documents and captured pages, not live social-media scraping.

Score: 5.7/10

Rubric security-score-v2.6; data completeness 0.65

Team transparency factors, points out of 10 (points)
Team identity
7.50 points
Disclosure
3.50 points
Development activity
5.50 points
FactorLevelEvidenceWeightPoints
Team identityL3supported0.457.50
Disclosure—not disclosed0.353.50
Development activityL2supported0.205.50

Track record: 8.0/10

Rubric security-score-v2.6; data completeness 0.80

Track record factors, points out of 10 (points)
Operating history
7.50 points
Incident record
10.00 points
Resilience
3.50 points
FactorLevelEvidenceWeightPoints
Operating history (computed)L3supported0.307.50
Incident record (computed)L4supported0.5010.00
Resilience—not disclosed0.203.50

45 citations · click markers to view evidence

🗳

Governance

5/10

Governance Analysis: Swell Liquid Staking

On the measured rswETH proxy-admin path, upgrades are controlled by a 4-of-7 Safe (a multisignature wallet) behind a 7-day timelock, which is a moderate setup; an rswETH staker-proxy upgrade, fund-moving permissions and the swETH controls are unresolved or not measured. The signers are only addresses; the available sources name no identities. The 7-day delay is shorter than the up-to-28-day rswETH withdrawal window, so users who want to exit before an upgrade lands may not be able to. Captures of the Snapshot voting space named in the docs (unverified publishers) show no proposals, and no source shows that a vote binds the Safe, so how changes are decided stays unconfirmed. In October 2026 Swell paused deposits and withdrawals; the cause, a compromised operational key, comes from Swell's X posts and press reports (unverified publishers) and stays unconfirmed. On-chain at block 26300146 the rswETH withdrawal, core and bot pauses are still switched on.

Governance Structure

  • Type: multisig-run with a timelock. The docs describe a Swell DAO with off-chain Snapshot voting, but on the measured rswETH path on-chain control sits with a Safe multisig and a timelock.
  • Voting mechanism: off-chain Snapshot. The docs do not state the voting weight rule or the quorum.
  • Proposal process: a draft goes to the forum for a minimum 7-day feedback period. A delegate or community manager then submits it to Snapshot. After a pass, "Swell DAO proceeds with the implementation". No document links a passed vote to a binding on-chain execution. The captured proposals page of the Snapshot space shows no proposals: (source reliability: unverified publisher). Captures of the Snapshot hub also return zero proposals and a null quorum for the space (source reliability: unverified publisher). What this means for actual governance activity is unconfirmed.
  • DAO scope today: the DAO's roles are phased. Contract and parameter development is listed under Phase 2, not as a current power.
  • Timelock: on Ethereum at block 26300146 the timelock 0xba82…073b owns the proxy admin of the rswETH contracts. It is a Compound-style timelock with a 604800-second (7-day) minimum delay, administered by the 4-of-7 Safe.
  • Emergency bypass: a pauser role is held by a Safe with threshold 1 of 5 owners and by a contract. It can pause core, operator, bot and withdrawal methods with zero delay. Unpausing needs the 4-of-7 Safe. The pause bypass is a benefit for emergencies. Its cost is that one signer can freeze exits. At block 26300146 on Ethereum the rswETH AccessControlManager reports bot, core and withdrawal methods as paused, and operator methods as not paused. Withdrawals through the protocol are therefore halted at that block, and the unpause needs the 4-of-7 Safe.
  • Vote concentration: not measurable. The measured vote file lists no venues. The Snapshot space named in the docs, swell-dao.eth , returns zero proposals in the unverified hub captures, so no votes, top voters or turnout can be computed from the supplied data. A voting row is omitted from the sidecar.

Admin Key Analysis

The 4-of-7 Safe is the main authority over rswETH upgrades. Its seven signer addresses are public, but no source links any of them to a named person or to Swell, so signer independence cannot be judged.

  • Upgrade authority (rswETH, Ethereum, block 26300146): the rswETH token, deposit manager, AccessControlManager and two more proxies share one proxy admin (0xd750…4bfd14). That admin is owned by the timelock, with a fastest upgrade delay of 604800 seconds.
  • Safe configuration: the Safe administering the timelock has 7 owners and threshold 4. The same Safe holds the unpauser role. It also shares signers with the 1-of-5 pauser Safe and a 2-of-6 Safe on the list; owner overlap is visible in the owner lists, but no signer is identified in any source. A Safe Transaction Service capture lists threshold 4 and seven owners for the same Safe, which agrees with the on-chain read. (source reliability: unverified publisher) An archived Balancer review lists this admin as 3/5. (source reliability: unverified publisher) That conflicts with the on-chain 4-of-7 and is treated as stale.
  • Roles on AccessControlManager: the timelock holds PLATFORM_ADMIN and DEFAULT_ADMIN_ROLE. It guards withdrawERC20, setDepositManager, setNodeOperatorRegistry, setRswETH, setRswEXIT and setSwellTreasury, each with a 7-day delay.
  • Fund-moving functions not resolved: 58 of 83 mapped privileged functions are marked unresolved. Examples include deposit-manager functions transferETHForEigenLayerDeposits, queueWithdrawals and withdrawERC20. Their role holders could not be read, so who controls them is not confirmed on-chain.
  • EigenPodManager: this is EigenLayer's contract. A 1-of-2 Safe upgrades it with no delay. That is an external dependency, not Swell's key.
  • October 2026 incident: Swell's own post says it paused deposits and withdrawals for swETH and rswETH. The Defiant (source reliability: unverified publisher) reports "compromised repricing keys" and says Swell announced no role revocation. A community incident report (source reliability: unverified publisher) says four malicious upgrade contracts had been deployed but none had been queued through Swell's multisig. This is a third party's statement and not a verified check of the timelock queue. Swell's own incident update on its X profile (source reliability: unverified publisher) gives its account of the cause: "On 21-22 October 2026, an attacker used a compromised operational key to inflate the swETH and rswETH exchange rates by about 6% and 5%." This is Swell's own account from an unverified publisher, so the cause stays unconfirmed; no full post-mortem is in the supplied sources. For the measured rswETH proxy-upgrade path, a 7-day timelock stands between an upgrade proposal and execution. Other fund-moving roles and the swETH controls remain unresolved, so this does not cover every way funds could move. The measured pause flags show withdrawals halted at block 26300146; whether the bot pause stopped rate changes for both products is unconfirmed.
  • swETH (v1 LST) contracts: the swETH token and its own AccessControlManager are not in the measured on-chain list. Their owners, thresholds and delays are not confirmed here. The claim that they match rswETH is unverified.
  • Reasonableness: a pause button and a timelocked upgrade path suit a staking protocol. The weaknesses are the undisclosed signer set, the 1-of-5 freeze power and unresolved fund-moving roles.

Upgradability

  • Pattern: EIP-1967 proxies (transparent-style admin). DepositManager, RswETH, AccessControlManager and EigenLayerManager each show an implementation and an admin on-chain.
  • Who upgrades (measured rswETH proxy-admin path): the 4-of-7 Safe, through the timelock, with a 7-day minimum delay. The upgradeStakerProxy path and the swETH proxies are not resolved or not measured.
  • Can upgrades change core invariants? Yes. A proxy upgrade can replace token and deposit logic, so a malicious or compromised upgrade could move funds. On the measured rswETH proxy-upgrade path, the 7-day delay and the Safe signers' review stand in the way; the swETH proxies are not measured.
  • Immutable parts: none identified. The Ethereum deposit contract and EigenLayer's contracts are external.
  • upgradeStakerProxy on EigenLayerManager is guarded by a role check whose holders could not be resolved, so its delay is unknown (not zero).

Parameter Controls

  • Timelocked setters: the six AccessControlManager setters listed above, with a 7-day delay.
  • Repricing parameters: the RswETH contract exposes maximumRepriceDifferencePercentage of 1% and minimumRepriceTime of 3600 seconds. This is a bound on each rate update. A bot key can still push roughly 1% per hour; the incident accounts (unverified publishers) describe rate inflation, and that cause stays unconfirmed. The oracle itself is assessed in the economic-security report.
  • Fee split: node operator and treasury reward percentages are each 5% on-chain (5e16). Who can change them was not resolved.
  • Harm potential: a bot key, if compromised, could inflate the displayed rate; a compromised or approved upgrade through the measured rswETH path could take effect only after 7 days; other fund-moving roles are unresolved.

Decentralization Assessment

On the measured rswETH path, a small multisig, not token holders, controls upgrades; fund-moving roles and the swETH controls are unresolved or not measured.

  • Spectrum: closer to "multisig-run with a timelock" than to DAO-run. Captures of the Snapshot space (unverified publishers) show no proposals, and nothing shows it binds the Safe.
  • Roadmap: the DAO docs lay out Phases 0.1, 1 and 2, with contract and parameter development only in Phase 2. No date is given.
  • Key person risk: one pauser signer can freeze deposits and withdrawals. A proxy upgrade on the measured rswETH path needs four signers plus 7 days; other fund-moving roles and the swETH controls are not resolved.
  • Token distribution: SWELL has a maximum supply of 10,000,000,000, with 25% to the team and 25% to fundraising, both vesting. Voter concentration and holder concentration were not measured.

Governance Risks

  • Single point of failure: the hot oracle/bot key and the 1-of-5 pauser. The rswETH withdrawal pause is still on at block 26300146.
  • Exit window versus delay: the withdrawal wait is up to 28 days. A 7-day timelock can be shorter than the withdrawal time. Requests that exceed the buffer of staking rewards and daily deposits can take longer. Smaller requests served from the buffer may finish sooner, and holders can also swap on a DEX.
  • Governance attack surface: captures of the Snapshot space (unverified publishers) show no proposals, and nothing shows it binds the Safe. If that holds, the exposure to vote-based attacks such as flash-loan voting or vote buying is limited. Bribery or compromise of the Safe signers is not assessed from the supplied sources, and it is the main exposure because the signers act at their own discretion.
  • Voter apathy: the captured Snapshot space (unverified publishers) shows no recorded proposals, and the hub returns a null quorum, so no participation can be measured from the supplied data.
  • Centralization creep: the three Safes share signers, and further powers (pause, unpause, upgrade) sit with overlapping people. Whether that overlap is the same persons is not known.

Data Completeness

  • Available: on-chain control state for the rswETH contracts on Ethereum (block 26300146), docs on the Snapshot process, and Swell's X post of 2026-10-23.
  • Missing: signer identities (a web search found only addresses); Snapshot vote concentration and quorum (the captured space shows no proposals and a null quorum); who executes any Snapshot decision; on-chain control of the swETH v1 contracts; holders of 58 unresolved privileged functions; a full post-mortem of the October 2026 incident beyond Swell's short incident update. The incident reports from The Defiant and the community come from publishers that could not be verified, so their detail is unconfirmed.
  • Audit evidence: audit digests were reviewed. Their governance findings are not used here; the on-chain state is current and the audits are dated 2022 to 2024.

Score: 5.0/10

Rubric security-score-v2.6; data completeness 0.75

Governance factors, points out of 10 (points)
Admin control
5.50 points
Change delay
5.50 points
Decision process
3.50 points
FactorLevelEvidenceWeightPoints
Admin controlL2supported0.455.50
Change delayL2supported0.305.50
Decision process—not disclosed0.253.50

22 citations · click markers to view evidence

🔗

Cross-Analysis

Cross-Risk Analysis: Swell Liquid Staking

Conclusion. Swell Liquid Staking (a DefiLlama record of about $28.5M TVL, all on Ethereum; the sources do not say whether it covers swETH, rswETH or both) is in an open incident. Swell paused deposits and withdrawals for swETH and rswETH, and the pause flags are confirmed on-chain. The cause, the size of any rate error and the amount of any loss are not confirmed by a reliable source. The reports together show three exposures: exits that stay closed with no reopening date, a rate-reporting role whose swETH holder is unknown, and no audit tied to the contracts running today. Holders should treat the position as impaired until Swell publishes a corrected rate, a root cause and a reopening plan.

Scope note: the protocol overview names swETH as the record's product, but the economic report says the mapping of the record to swETH, rswETH or both is not stated . This report therefore keeps swETH as the main product and treats the $28.5M as the value of the record. rswETH (liquid restaking) is a separate family product, covered here only where it shares code, keys or incident exposure with swETH . Most on-chain control data (pause and unpause authorities, timelock, repricing caps) is measured for rswETH only. For swETH, direct reads cover the token, its proxy admin and its pause flags, and the holders of its pause and reprice roles were not read .

Top Risks for a Holder

The three risks below are ranked by size of possible loss and by how much of the evidence is established. Figures from press reports (unverified publisher) are marked as unconfirmed.

swETH rate in ETH: contract read vs press-reported clean rate and pool price (ETH)
Contract rate, block 26300482
1.203 ETH
Clean rate, press report unverified
1.133 ETH
Pool price 23 Oct, press unverified
1.089 ETH

The contract rate is an on-chain read. The other two bars come from one press article (source reliability: unverified publisher). The clean rate is an analyst's claim and the pool price is a dated snapshot, not a current exit price.

1. Exits stay closed, and the rate may be reset (swETH and rswETH)

  • What goes wrong. Withdrawals and deposits are paused for swETH and rswETH. Swell says the next step is to "return NAV to the correct value" before resuming (Swell X post, source reliability: unverified publisher). The swETH contract rate read 1.203 ETH at block 26300482. Swell has not published a corrected rate or reopening date. Press analysts put the last clean rate at 1.133 (chart above, unconfirmed), which would mean a fall of about 5.8% from 1.203 (my arithmetic). The same article put swETH at about 1.089 WETH in one Uniswap pool on 23 October, and rswETH at about 1.029 WETH on Curve against a contract rate of 1.1377 (source reliability: unverified publisher).
  • Why it is cross-domain. The pause is a governance control. For rswETH, a pauser Safe with a 1-of-5 threshold can pause with no delay, and unpausing needs the 4-of-7 Safe . The holders of the swETH pause and unpause roles were not read. For swETH only, the Hexens audit of the February 2024 swETH update recorded as acknowledged that pauseWithdrawals had no automatic resume . Whether the current swETH deployment differs was not compared , and the sources do not show that rswETH shares that behaviour. Secondary liquidity data is thin. The rswETH depth probe found zero depth at 2% price impact . swETH depth is not measured (source reliability: unverified publisher for the pool figure).
  • Magnitude. Swell's own exit route is closed for both tokens. The recorded TVL is $28.5M . The sources do not say whether that figure covers swETH, rswETH or both, and press reports put each at about $29M (source reliability: unverified publisher), so the exposed total across both tokens may be larger. If the press-reported clean rate were right and the whole $28.5M were swETH priced at the inflated rate, the reset would cut displayed value by about $1.7M (my arithmetic). Whether principal is impaired depends on backing relative to liabilities and on the redemption terms after the reset. The economic report states the backing-versus-liabilities ratio is not measured and that absence of a cited shortfall is not proof of full backing . Swell's statement that no ETH left the deposit or withdrawal contracts, and an analyst's remark that the 2,513 ETH figure is "NOT confirmed loss", are both unconfirmed (source reliability: unverified publisher).
  • Likelihood. High that exits stay closed until Swell acts: the on-chain pause flags are set and no reopening date exists. The duration is unknown. The direction and size of the correction are unconfirmed: Swell's statement says only that NAV must be returned to the correct value. The published corrected rate and a backing measurement would resolve both. The press-reported clean rate, if right, would imply a fall.

2. Misuse of the rate-reporting (bot) role after reopening (swETH, rswETH, and any lender using the rate)

  • What goes wrong. Reports allege a compromised operational key raised the rates in steps of 1% an hour with no ETH behind them (source reliability: unverified publisher). That account is an inference by a third party and is unconfirmed. A second press source says no role-revocation event had appeared (source reliability: unverified publisher). If the role holder or key is not changed before the unpause, the same path stays open.
  • Why it is cross-domain. Governance cannot say who holds the repricing role: 58 of 83 mapped privileged functions are unresolved . The economic report says the holder of the swETH repricing key and any current on-chain bound for swETH are open gaps . The code report separates custody from code: a key compromise would be an operational matter, not a finding about audited code .
  • Magnitude. Up to the redemption value of the pool if an inflated rate were honoured on exit. For rswETH, where the 1% per step cap is measured, repeated steps compound, so the cap slows an increase without bounding its total . A press report on swETH only says its withdrawal contract pays the lower of the request-time and processing-time rate, which would limit this for swETH if it holds (source reliability: unverified publisher). The economic report says the rule's current scope for rswETH is not confirmed , so it is not counted as a stop for rswETH. For swETH, the cap, the total possible change and the role holder are unresolved, so its magnitude is not estimated. Lenders that price off the contract rate are also exposed. The only sized case is a Morpho market with about 34 ETH to borrow .
  • Likelihood. Assessment: medium, low confidence. Exposure: a bot role submits the reserve figure for repricing , and a press source says no role-revocation event had appeared (above), so the same key could still hold the role at reopening. Checks that lower it: the audited design validates the bot's figure against a proof-of-reserves oracle , which is the audited version and not shown to be current, plus the rswETH step cap and the pause. Prerequisite: control of the role holder's key. The probability of misuse is not measured, and abuse in October is unconfirmed. Facts that would move the band: the swETH role holder, any current swETH bound, whether the proof-of-reserves check is live, and a key rotation notice.

3. Compromised signers or infrastructure reaching the upgrade path (swETH, rswETH)

  • What goes wrong. A press report says four malicious upgrade contracts able to take about 2,500 ETH were deployed but not queued through the multisig (source reliability: unverified publisher). On the measured rswETH path, an upgrade needs the 4-of-7 Safe plus a 7-day timelock. The swETH proxy admin is owned by the same timelock address, so the same delay probably applies .
  • Why it is cross-domain. For rswETH, the three Safes share signers, and no source ties any signer to a named person . The social report records an unexplained 22 October repository commit titled "ci: zz-w22-s08 secret dump (org-secret proof)" and notes that researchers raised a possible Railway cloud-platform link, which Swell has not confirmed (source reliability: unverified publisher). The sources do not connect the commit to any signer key.
  • Timelock versus exit time (when withdrawals are open). rswETH withdrawals take up to 28 days , so some rswETH exits could outlast a 7-day timelock. swETH withdrawals typically take 1-7 days, with delays of up to 12 days or more . Smaller requests served from the buffer may finish sooner, and holders can swap on a DEX. The pause is a separate impediment to exit now.
  • Magnitude. Loss of pooled ETH reachable by the upgraded contracts. The press figure of about 2,500 ETH would be roughly one fifth of rswETH's supply at its contract rate. That is arithmetic on an unconfirmed input (10,654 tokens at 1.1377, about 12,100 ETH) . The second figure, 2,513 ETH, is called "NOT confirmed loss" by the same source and describes ETH held.
  • Likelihood. Low. The queue step has a 7-day public window and nothing was reported as queued. Signer identities, signer key custody and any monitoring of the timelock queue are not in the sources, and they are the facts that would change this band.

Combined Attack Vectors

Vector 1: Rate inflation, then redemption or lending extraction

  1. An actor with the repricing (bot) role raises the rate 1% an hour. A third-party report alleges this began at 23:02 UTC; the last rswETH reprice on-chain was at 04:25 UTC on 22 October .
  2. For rswETH, each step stays under the measured on-chain cap of 1% with a one-hour minimum, so the contract accepts it. No bound is measured for swETH, so its exposure per step is unknown .
  3. The inflated rate lets the actor mint fewer tokens per ETH, redeem more ETH per token, or borrow against inflated collateral in a market that reads the contract rate.
  4. Stops: the withdrawal queue (no flash-loan loop) and the pause. A press report on swETH describes a lower-of-two-rates rule (source reliability: unverified publisher), but its scope for rswETH is not confirmed, so it is not counted as a stop there. The rswETH pause can be set by one signer of a 1-of-5 Safe; the swETH pause holders were not read.
  • Domains. Governance (role holders unresolved), economic (rate-based pricing), code (the audited design trusts a bot role; Hexens fixed a medium where a BOT-role holder could drain the withdrawal queue by reentrancy ), and social (cause unconfirmed).
  • Likelihood. Medium (assessment, low confidence): see Risk 2 for the checks, prerequisites and missing facts; the swETH holder and bound are unresolved. Impact. Severe for rswETH if repeated capped steps compound and an inflated rate is honoured on exit. For swETH the impact is unknown until its bound is read.

Vector 2: Infrastructure or signer compromise leading to a timelocked malicious upgrade

This vector is a hypothesis. The Railway link is unconfirmed by Swell, and no source ties any signer to a person or to a breach, so none of the first steps is shown to have happened.

  1. Hypothesis: a breach of shared cloud or CI secrets exposes operational keys.
  2. The actor pre-deploys malicious implementations (press report: four).
  3. The actor needs four of seven signers on the rswETH admin Safe (signers shared across three Safes) to queue an upgrade.
  4. The 7-day delay starts. Swell or the community must notice and cancel.
  5. If not cancelled, the upgrade executes and funds can move. Holders may be unable to leave within 7 days if exits are paused or queued longer.
  • Domains. Social (unconfirmed breach lead, unexplained commit, unidentified signers), governance (signer overlap, delay versus exit time), code (no audit tied to deployed addresses; source collected for 4 of 15 listed addresses), economic (exit queue length).
  • Likelihood. Low, since no breach path or signer exposure is verified. Impact. Catastrophic if it succeeds.

Vector 3: Reopen-and-run

  1. The unpausing role reopens. For rswETH this is the 4-of-7 Safe with no delay . The swETH unpauser was not read. The sources show no notice period.
  2. NAV is corrected. Direction and size are not published; the press-reported clean rate would imply a fall.
  3. Holders race to exit at whichever rate is honoured, into a buffer that large requests can exceed .
  4. Those who cannot queue sell on DEXs. rswETH has zero measured depth at 2% impact, and a $100,000 rswETH probe moved the price 19.41% . swETH depth is unmeasured, and the one pool figure of about $143,000 is a dated press snapshot of a single pool.
  • Domains. Governance (unpause timing), economic (thin liquidity, queue buffer), social (no reopening plan or post-mortem).
  • Likelihood. Conditional scenario, medium for rswETH if exits reopen at a lower corrected rate. It rests on zero measured rswETH depth at 2% impact and a no-delay rswETH unpause. The reopening terms and swETH depth are unknown, so the swETH band is unset. Impact. Severe for late exiters; moderate for the protocol overall.

Vector 4: Long freeze with no shown holder recourse

  1. Whoever holds the pause role keeps withdrawals paused. For rswETH, one signer of a 1-of-5 Safe can set the pause. The swETH pause holders were not read.
  2. For swETH only: the audited February 2024 swETH update had no automatic resume (acknowledged). Whether the current swETH deployment differs was not compared, and the sources do not show that rswETH shares the behaviour, so the long-freeze path for rswETH rests on the pause authority alone.
  3. A capture of the Snapshot space named in the docs shows no proposals (source reliability: unverified publisher). No source shows that a vote binds the Safe, so how changes are decided is unconfirmed. This does not prove holders have no other recourse.
  • Domains. Code (acknowledged finding in the audited version), governance (pause authority, no source shows a vote binds the Safe), social (no communication after 23 October in the sources).
  • Likelihood. The duration of the current freeze is unknown, since no reopening date exists; low that it is malicious. Impact. Severe for liquidity, moderate for solvency.

Systemic Risk Factors

  • Dependencies. The dependency graph lists only WETH as backing (share 1.0, status unresolved), Ethereum as chain, and no scored protocol dependency. Its warnings list is empty (collected 2026-10-29) . The graph does not list the dependencies the economic report names: EigenLayer (rswETH), the Chainlink proof-of-reserve oracle, permissioned node operators, and LayerZero/Nucleus bridging. Route these to the dependency-risk owner for reassessment.
  • EigenLayer (rswETH). The EigenPodManager upgrade path is a 1-of-2 Safe with no delay and is not Swell's key .
  • Bridges. A Swell X post of 19 April 2026 says it paused LayerZero bridging after the Kelp rsETH incident . A later post says DVN settings rose from 1/1 to 3/3 . Both are self-reported X posts (source reliability: unverified publisher) and are unconfirmed. All TVL is on Ethereum, so only bridged holders are exposed.
  • Correlated exposure. swETH and rswETH share a design, which supports correlated exposure. The rswETH docs describe it as remaining "exactly the same as swETH" apart from EigenLayer . Both were paused in October. Key overlap between them is unresolved, because the swETH pause and reprice role holders were not read. rswETH also carries EigenLayer-specific dependencies that swETH does not.
  • Lending venues. A third-party incident report that its author said was AI-written and not fully checked says Aave, Spark and Fluid did not accept either token as of 21 October (source reliability: unverified publisher). The report covers three venues on one date, so it does not establish the extent of lending contagion or the exits available.
  • Macro. An ETH price fall or a wider liquid-staking confidence shock would hit already thin secondary markets. No sourced figure sizes it.

Risk Amplifiers

  • Privileged rate role + unresolved holder + no measured swETH bound. The rate can move on a role whose swETH holder is unknown, and no on-chain step cap is confirmed for swETH.
  • Single-signer pause (rswETH) + no source showing that a Snapshot vote binds the Safe. For rswETH, one signer can freeze exits and unpausing needs the 4-of-7 Safe. The Hexens no-auto-resume finding concerns the audited swETH update only. The code report says no source shows that audited access control manager is now deployed for swETH or rswETH , so it is not carried over to rswETH, and whether swETH keeps it was not compared.
  • 7-day delay versus up-to-28-day rswETH exits. Some rswETH exits could outlast the upgrade timelock when withdrawals are open. swETH exits typically take 1-7 days.
  • Unlinked audits + incident after the last audit. The most recent swETH audit in the workspace is from February 2024, and the October 2026 incident post-dates every audit. The comparison of deployed code with audited versions was not made: 100.0% of 4,770 lines in 22 verified files are marked "unknown", which means not compared, not a measured mismatch .
  • Rate-based TVL. The headline $28.5M may be overstated if priced off the contract rate. The sources do not say how DefiLlama prices the tokens.
  • Thin disclosure. No post-mortem, reopening date, signer identities or key rotation notice appear in the sources, so holders cannot price any of the above.

Risk Mitigators

  • Pause used. Swell announced the pause on 23 October (X post, source reliability: unverified publisher). The swETH withdrawal pause is confirmed on-chain at block 26300483 . The rswETH pauses are confirmed at block 26300146 .
  • Per-update rate cap (rswETH only). 1% per step and a one-hour minimum limit how fast the rate can rise. No cap is measured for swETH.
  • Withdrawal design. Queued exits block same-block loops. A press report on swETH describes a lower-of-two-rates rule (source reliability: unverified publisher); its scope for rswETH is not confirmed.
  • Timelock on upgrades (measured for rswETH, inferred for swETH). A 7-day delay on a 4-of-7 Safe. Nothing was reported as queued (press, unverified publisher).
  • Audit history. The Hexens February 2024 review of the swETH update found 1 medium and 7 low/informational findings, all fixed or acknowledged. Its link to deployed addresses is not confirmed. The March 2022 SoftStack/Chainsulting audit of an earlier swETH version had one critical finding (initialize not protected), which the team acknowledged . The team's update says the proxy is guarded by the initializer modifier . The code report says this was not checked against code , so it is not a demonstrated current vulnerability. Codespect pages list a repricing update (April 2025) and a deposit-bot report (July 2026), read from summary pages only (source reliability: unverified publisher).
  • Backing statements. Swell says no ETH left the contracts and "All funds are safe". These are first-party X posts (source reliability: unverified publisher) and are not independently confirmed.
  • Hacks dataset. The DefiLlama hacks list for the family is empty, but it predates the incident and says nothing about it.
  • No bad debt cited. Swell runs no lending, so there is no protocol liquidation cascade.

Contradictions & Inconsistencies

  • Cause of the incident. The governance report attributes the compromised operational key to Swell's X posts and press reports (unverified publishers) and calls the cause unconfirmed. The social report likewise marks the key account unverified. The overview and economic reports say Swell has not confirmed a cause and that no first-party post-mortem was found . The X capture shows 660 followers and "Joined May 2025", so the account's official status is itself unproven. This report treats the cause as unconfirmed.
  • Incident record versus event. The social report scores the incident record at the top level (L4, 10.00 points), based on an empty hacks dataset. The same report documents an unresolved October 2026 pause and an April 2026 bridge pause that rests on an unverified-publisher X post and is unconfirmed. The score is mechanical and does not reflect the live event.
  • Score spread. Code security is 6.0/10 (capped because audits are not linked to deployed code). Economic security is 3.9/10 (solvency at the lowest level because exits are halted). Governance is 5.0/10 (its decision-process factor is not disclosed, which lowers data completeness to 0.75) and team transparency 5.7/10. The code score measures audit coverage and findings. It does not measure operational key custody, which the code report places outside audited-code findings.
  • Oracle bound. The economic report says no current bound is known for swETH, while the rate caps are measured for rswETH only. A press source describes both rates as moved by about 6% and 5%. The size and mechanism for swETH are unconfirmed.
  • Timelock reach. The overview infers the swETH timelock from a shared owner address. The governance report says swETH owners and delays are not confirmed and that the match to rswETH is unverified.
  • Record scope. The overview says the record's product is swETH. The economic report says the mapping of the record to swETH, rswETH or both is not stated. This report uses the $28.5M as the record's value and does not assign it to one token.
  • Audit attribution. Swell's docs name Sigma Prime and Codespect (and, via a blog, Cyfrin and Hexens for Barracuda). The code report finds only the Hexens 2024 review readable and tied by name to swETH. Sigma Prime and Codespect PDFs were not read, so nothing is shown to contradict the docs, but none of that coverage is counted.
  • Pause scope. The governance report says the measured pause flags show withdrawals halted at block 26300146, and that whether the bot pause stopped rate changes for both products is unconfirmed. A press source says pausing the bot path is not key revocation, and no revocation is visible.

Overall Risk Profile

Swell's composite risk is high at the research date (2026-10-29). The protocol is small, its swETH value rests on an off-chain bot reporting an exchange rate, and the swETH role holders for that rate and for pausing were not read. The established facts are the on-chain pause, a contract rate Swell says must be returned to a correct value, thin rswETH liquidity and audits that are not tied to deployed code. The main exit-risk scenario is conditional. If exits reopen at a corrected rate below the displayed one while secondary liquidity stays thin, late exiters could face slippage. The pause is established. The reopening terms, the corrected rate and swETH depth are unknown, so this is a scenario and not a forecast. A press-reported clean rate would imply a fall of about 5.8% for swETH, but that figure is unconfirmed. The worst case is renewed misuse of the rate role, or a signer-level compromise that upgrades contracts to drain funds. Neither is established by the sources. No peer data is in the sources, so no sourced comparison with other liquid-staking protocols is made.

Data Gaps

  • Collection of GitHub repositories stopped at the collection limit; some material may be missing.

  • Current deployment addresses could not be confirmed for some audited contracts.

  • Source summaries cover only the first 80000 characters of 3 document(s); later content was not assessed.

  • No first-party post-mortem, key rotation notice, corrected NAV or reopening date. The cause and loss size cannot be set, and the clean-rate figure rests on one unverified analyst.

  • Role holders of swETH reprice, pause and unpause functions, and the swETH Deposit Manager and registry admins, are unread. This limits Vectors 1, 3 and 4.

  • Signer identities and key custody of the three rswETH Safes are not established. This limits Vector 2.

  • The delay and holders for the rswETH staker-proxy upgrade function and other fund-moving permissions are unresolved, so the 7-day timelock may not cover every path to funds. This limits Vector 2.

  • The timelock queue was not read directly, and the "none queued" claim is from press.

  • No backing-versus-liabilities measurement, swETH DEX depth, holder concentration, or APY and yield history.

  • Whether the current swETH deployment differs from the audited February 2024 version, including the no-auto-resume behaviour, was not compared.

  • Sigma Prime and Codespect audit text, and current bug-bounty status, are not read.

  • How DefiLlama prices swETH and rswETH, and which product the $28.5M covers, is not stated.

  • The dependency graph and the economic report list different dependency sets, so correlated-failure risk could not be tested.

60 citations · click markers to view evidence

Preview3 · Default / Rows