The rsETH Bridge Drain: LayerZero DVN Compromise, Aave Contagion, and a CROPS Audit

Published: April 20, 2026

tl;dr

• An attacker drained 116,500 rsETH (~$290–292M) from Kelp DAO’s LayerZero-powered Unichain → Ethereum bridge on April 18, 2026 at 17:35 UTC — the largest crypto exploit of 2026 to date, eclipsing the $285M Drift drain from April 1.
• The bug was not in any smart contract — it was a Decentralized Verifier Network (DVN) configuration set to “1-of-1,” combined with a multi-stage infrastructure attack: two op-geth RPC nodes were compromised with malicious binaries, then a DDoS forced LayerZero’s verifier to fail over to the poisoned nodes, which forged a valid cross-chain message.
• LayerZero attributed the operation to North Korea’s Lazarus Group (TraderTraitor subgroup) and committed to stop signing messages for any 1-of-1 application; the attacker was pre-funded via Tornado Cash ~10 hours before the exploit (per ZachXBT).
• Kelp’s emergency pause hit 46 minutes later and blocked two follow-on attempts of ~40,000 rsETH each (each worth roughly $100M per CoinDesk) — without it, the combined drain would have been on the order of $490M.
• Aave is the contagion epicenter: the attacker deposited 89,567 rsETH (~$221M) as collateral and borrowed against it across V3 Ethereum and Arbitrum. Aave’s bad debt sits between $123.7M (uniform socialization) and $230.1M (L2-isolated) depending on Kelp’s redemption design.
• Aave’s response was textbook — Guardian initiated freezes at 18:52 UTC and completed them across all 11 V3 deployments by 19:00 UTC (LTV → 0). WETH borrow caps and rate adjustments followed within 24 hours, and the V4 Security Council disabled supply/borrow on Ethereum. No Aave contract vulnerability was involved.
• DeFi-wide TVL fell ~$13B over 48 hours (CoinDesk), AAVE traded down 16–18%, and contagion forced freezes or pauses at Compound, Fluid, SparkLend, Euler, Lido Earn, Morpho, and Kamino. Aave’s own TVL dropped $6–8.5B depending on snapshot.
• The dispute over blame is unresolved: Kelp says LayerZero’s own quickstart and default GitHub configuration ship 1-of-1 setups (used by ~40% of LayerZero protocols); LayerZero says it sent direct guidance to upgrade. Independent researchers (banteg, Zach Rynes) sided with Kelp on the defaults question.
• Through the CROPS lens (Censorship resistance, Resistance to capture, Open source, Privacy, Security), this incident is primarily a Resistance-to-Capture and Security failure — every layer that should have provided redundancy collapsed to a single point: one DVN, one verifier operator, one set of RPC endpoints, and one bridge adapter custodying assets for 20+ chains.

Table of Contents

1. Executive Summary
2. The Three Systems Involved: rsETH, LayerZero V2, Kelp’s Bridge
3. Anatomy of the Exploit
4. Sequence of Events: April 18 – April 20, 2026
5. Aave Impact and Mitigation Efforts
6. Cross-Protocol Contagion
7. The Kelp / LayerZero Configuration Dispute
8. Recovery Options on the Table
9. Accountability: A CROPS Audit
10. Sources

1. Executive Summary

On April 18, 2026 at 17:35 UTC (Ethereum block 24,908,285), an attacker forged a cross-chain message on Kelp DAO’s LayerZero V2 Unichain → Ethereum rsETH bridge and minted 116,500 rsETH (~$290–292M) to an attacker-controlled address. The attacker then dispersed the unbacked rsETH across Aave V3 (Ethereum and Arbitrum), Compound V3, and Euler, posting it as collateral and borrowing approximately $236M in real assets — primarily WETH.

The exploit succeeded because three independent failure modes lined up:

1. A 1-of-1 DVN configuration on Kelp’s bridge, meaning a single LayerZero Labs verifier signed off on cross-chain messages with no redundancy.
2. Two compromised op-geth RPC nodes the LayerZero Labs DVN relied on, with attacker-swapped binaries returning forged data to the verifier while serving truthful responses to other clients (including LayerZero’s own monitoring).
3. A targeted DDoS against the uncompromised RPC nodes that forced failover to the poisoned ones during the attack window (~10:20–11:40 AM PT / 17:20–18:40 UTC).

Kelp’s emergency pause activated 46 minutes after the initial breach and blocked two follow-on attempts targeting ~40,000 rsETH each (CoinDesk values each at roughly $100M). Without that pause, the combined drain would have been on the order of ~$490M.

The bug was not in any smart contract. The Aave protocol itself never failed — the contagion is a collateral integrity failure propagated through interconnected lending markets that accepted rsETH as collateral.

The remainder of this report walks through the technical mechanics, the minute-by-minute timeline, Aave’s mitigation playbook, the cross-protocol fallout, the unresolved blame dispute between Kelp and LayerZero, and a CROPS-framework audit of where each party stands accountable.

2. The Three Systems Involved: rsETH, LayerZero V2, Kelp’s Bridge

2.1 rsETH

rsETH is Kelp DAO’s liquid restaking token (LRT) representing restaked ETH positions. Like other LRTs (ezETH, weETH), it sits at the intersection of staking, restaking, and DeFi — circulated as collateral across lending markets and yield strategies on Ethereum mainnet plus 20+ Layer 2 and alternative L1 networks.

Because rsETH is backed by ETH on Ethereum mainnet but circulates on remote chains, it relies on a lock-and-mint bridge model: rsETH on a remote chain is a claim against an adapter contract on Ethereum that is supposed to hold an equivalent amount in custody. Breaking the invariant that remote-chain rsETH ≤ adapter-held rsETH is the entire game.

2.2 LayerZero V2 and the DVN

LayerZero V2 is a generalized cross-chain messaging protocol. Its core innovation is the Decentralized Verifier Network (DVN) model: instead of a single fixed validator set, each application chooses an N-of-M verifier configuration. A message is treated as valid only when the configured threshold of DVNs independently confirms it.

The model is intentionally pluralistic: applications can pick LayerZero Labs’ DVN, Polyhedra, Google Cloud, Nethermind, or any combination, and require unanimous or threshold consensus.

The security guarantee, however, is only as strong as the configuration the application chooses. A 1-of-1 DVN setup has fault tolerance of zero — compromise the one verifier, forge any message. A 2-of-3 setup requires compromising two independent verifier networks simultaneously. A 5-of-9 raises the bar further.

2.3 Kelp’s Bridge Adapter

Kelp’s rsETH bridge is built on LayerZero V2’s OFT (Omnichain Fungible Token) standard. The Ethereum-side adapter custodies rsETH; remote-chain OFT contracts mint and burn rsETH against incoming LayerZero messages.

Crucially, Kelp configured its Unichain → Ethereum route as 1-of-1 with LayerZero Labs as the sole verifier. That meant a single message signed by a single DVN, sourced from a single set of RPC nodes, could trigger the Ethereum adapter to release rsETH.

3. Anatomy of the Exploit

3.1 The Bridge Invariant That Broke

Per Aave’s incident report, the bridge state immediately around the exploit:

The adapter went from holding ~116,723 rsETH to ~223 rsETH — effectively drained. Subsequent inflows partially refilled it to 40,373, but with 152,577 rsETH still claiming backing across remote chains, the shortfall is 112,204 rsETH, roughly 73.5% of bridged supply.

3.2 The Multi-Stage Attack

The exploit is best understood as three stages stacked on top of a single configuration weakness:

Stage 1 — Reconnaissance and pre-funding. The attacker funded the operating wallet through Tornado Cash approximately 10 hours before the exploit (per ZachXBT’s on-chain trace). They had also identified the specific RPC endpoints the LayerZero Labs DVN was querying for Unichain state.

Stage 2 — RPC node compromise. The attacker gained access to two op-geth nodes serving the DVN, running on independent clusters with no direct connection to each other. Attacker-controlled binaries were swapped in. The compromised nodes were programmed to return forged Unichain state only to the DVN’s IPs while serving truthful responses to all other clients — including LayerZero’s own Scan service and internal observability infrastructure. This stealth design is what kept the compromise invisible to LayerZero’s monitoring.

Stage 3 — DDoS-driven failover. LayerZero’s verifier did not query only the two compromised nodes — it also reached out to additional RPC providers. The attacker ran a distributed denial-of-service attack against the uncompromised nodes between ~10:20 AM and 11:40 AM PT (17:20–18:40 UTC), forcing the DVN’s failover logic to route to the poisoned endpoints.

Once the verifier was reading forged data, the attacker submitted a cross-chain message claiming rsETH had been locked on Unichain. The DVN — the only verifier in the 1-of-1 set — confirmed it. Kelp’s Ethereum adapter released 116,500 rsETH to the attacker-controlled address.

3.3 Failed Follow-Ups and the 46-Minute Window

The attacker attempted two more drains targeting ~40,000 rsETH each in rapid succession after the initial transaction. Kelp DAO’s pauser multisig activated emergency pauses 46 minutes after the first exploit transaction, freezing deposits, withdrawals, oracle functions, and the rsETH token across mainnet and several L2s. Both follow-on attempts failed.

Without the pause, the combined drain (the original 116,500 rsETH plus two further ~40,000 rsETH attempts that CoinDesk valued at roughly $100M each) would have been on the order of ~$490M.

For scale: the 116,500 rsETH actually drained represents approximately 18% of rsETH’s ~630,000 circulating supply.

3.4 Attribution

LayerZero’s post-mortem attributes the operation to DPRK’s Lazarus Group, specifically the TraderTraitor subgroup, citing “preliminary indicators” consistent with prior North Korea-linked exchange and bridge attacks. The combination of pre-funding via Tornado Cash, sophisticated infrastructure compromise (binary-level RPC tampering with stealth payload), and coordinated DDoS is consistent with TraderTraitor’s pattern.

The attribution remains preliminary. LayerZero says it is cooperating with law enforcement on fund tracing.

4. Sequence of Events: April 18 – April 20, 2026

All times UTC unless noted.

5. Aave Impact and Mitigation Efforts

5.1 Attacker Positions on Aave

The attacker treated Aave V3 as the primary monetization venue. Per Aave’s incident report:

Aave V3 Ethereum Core:

• Address 1: 53,000 rsETH collateral → 52,460 WETH borrowed
• Address 2: 400 rsETH collateral → 394 WETH borrowed

Aave V3 Arbitrum (6 addresses):

• Range: 770–12,574 rsETH supplied per address
• Total: ~36,167 rsETH collateral → ~25,455 WETH borrowed

Total deposited on Aave: 89,567 rsETH (~$221.39M) out of the 116,500 rsETH stolen — roughly 77% of the haul ended up in Aave alone. The remainder went to Compound V3 and Euler.

5.2 Bad Debt Scenarios

Reading the two numbers correctly. The loss is a single fixed fact: the bridge adapter is missing 112,204 rsETH of backing. That number does not change between scenarios. What changes is the policy choice about who absorbs that shortfall — and rsETH is fungible by branding but not by mechanics. Mainnet rsETH is a direct claim on Kelp’s Ethereum staking contracts (still fully backed). L2 rsETH is an OFT-bridged claim against the adapter that was drained. So the same 112,204 rsETH shortfall maps to two very different Aave exposures depending on whether Kelp treats all rsETH as mutually backed (Scenario 1) or honors the underlying mechanics (Scenario 2). The “$123.7M vs $230.1M” range is not Aave guessing the loss — it is Aave modeling two outcomes of a Kelp governance decision that has not yet happened.

Aave’s exposure depends on how Kelp DAO ultimately structures rsETH redemption. The incident report models two scenarios:

Scenario 1 — Uniform loss socialization (15.12% haircut on all rsETH):

Scenario 2 — L2-isolated losses (73.54% haircut on bridged rsETH only):

The choice between these scenarios is external to Aave — it depends entirely on whether Kelp DAO opts to socialize losses across the full rsETH holder base or concentrate them on remote-chain holders whose backing was the portion drained.

The principled answer is Scenario 2 — but with a caveat

Bridge risk is on the bridge user. That is the only consistent rule that produces correct long-term incentives in DeFi:

• Mainnet rsETH was never bridged. Its backing is intact on Ethereum. Forcing mainnet holders to absorb the bridge failure would mean people who took no bridge risk subsidize people who did. That breaks the price signal that should make markets demand better bridge configurations.
• Socialization rewards bad bridge configs. If issuers can ship a 1-of-1 DVN, custody $290M, lose it, and have the loss spread evenly across all holders, they have no incentive to ship 3-of-5. The cheapest configuration wins. That is the world the broader CROPS audit (§9) is trying to prevent.
• It is consistent with the CROPS framing. Ethereum mainnet is the sanctuary. Bridges are not. Treating bridged claims as fully fungible with mainnet claims erases the distinction the framework exists to preserve.
• It is what the mechanics actually say. OFT-bridged rsETH on Mantle is, literally, a claim against an Ethereum adapter that no longer holds the assets. Calling it equivalent to mainnet rsETH is a marketing choice, not a mechanical fact.

The caveat: the L2 holders did not choose the bridge configuration — Kelp did. “Bridge risk is on the user” is the right rule for loss allocation (mechanically, the L2 holders are the ones holding the broken claim). But it is the wrong stopping point for responsibility. The user took the bridge risk; they did not take the “1-of-1 DVN configured by Kelp against LayerZero’s stated best practice” risk. Those are different.

The clean structure is therefore:

1. Loss falls on bridged supply (Scenario 2 mechanics). L2 rsETH absorbs the haircut; mainnet rsETH stays whole.
2. Restitution flows from Kelp (and where applicable, LayerZero). Kelp DAO treasury, future protocol revenue, KERNEL token issuance, and any LayerZero contribution are deployed to make L2 holders progressively whole. The L2 user is the first-loss tranche, but they are not the only tranche.
3. Aave models the worst case (Scenario 2, $230.1M) for solvency planning even if a Kelp-led recovery later compresses the realized loss.

Socialization (Scenario 1) is the politically easy answer — spread the pain — but it is the wrong precedent. The next 1-of-1 bridge that loses $500M will cite this case as the reason to socialize again, and the cost of insecure bridge defaults will continue to be paid by holders who never touched the bridge.

5.3 Defensive Actions Taken

Phase 1 — Immediate freeze (April 18, 18:52–19:00 UTC):

• Aave Protocol Guardian froze rsETH and wrsETH across all 11 V3 deployments where listed (Ethereum, Prime, zkSync Era, MegaETH, Mantle, Base, Plasma, Arbitrum, Avalanche, Ink, Linea), setting LTV to 0.
• Aave V4 Protocol Security Council disabled supply and borrow activity via two transactions on Ethereum.

Phase 2 — WETH market protection (April 19, 02:28 – April 20, 05:00):

• Precautionary WETH borrowing freeze on Core, Prime, Arbitrum, Base, Mantle, and Linea.
• WETH interest rate adjustments via Risk Steward to stabilize utilization and slow attacker borrows.
• Core WETH parameters retuned: Slope 1 to 2%, Slope 2 to 3%, optimal utilization to 94%.

Phase 3 — Umbrella decision pending:

• Aave’s Umbrella safety module holds 23,507.63 WETH (~$54.06M), of which 18,922 aWETH (80.5%) is in cooldown — meaning a meaningful share is already positioned to exit unless action is taken.
• Recommendation: immediate preventative pause of the WETH Umbrella module. This prevents capital flight while preserving manual governance control over any coverage deployment.
• Under Scenario 1, the Umbrella could offset roughly 59% of the $91.79M Ethereum Core WETH bad debt.

5.4 The WETH Liquidity Crisis

A second-order problem emerged immediately after the freezes: all five major Aave WETH reserves (Ethereum, Arbitrum, Base, Linea, Mantle) hit 100% utilization with idle balances under $20. This means:

• Liquidators cannot receive underlying WETH on liquidation; they receive aWETH instead, which they then must unwind separately. This slows liquidation throughput.
• The first liquidations of attacker positions (or any other rsETH-collateralized positions) would trigger at very small WETH price drops — 0.77–1.77% on Base/Arbitrum, 22% on Mantle (the most resilient market).
• A WETH price wobble that would normally be a non-event becomes a forced liquidation cascade.

5.5 DAO Financial Position

Aave’s ability to absorb bad debt directly is constrained by treasury composition:

For context, Aave generated $145M in revenue in 2025 and $16M net income YTD 2026. A $123.7M bad-debt outcome (Scenario 1) is roughly within reach via combined treasury deployment + Umbrella + future revenue. A $230.1M outcome (Scenario 2) would require either AAVE token issuance, a Safety Module slashing event, or an external recovery contribution from Kelp / LayerZero.

6. Cross-Protocol Contagion

The exploit is a case study in DeFi composability cutting both ways. rsETH was integrated as collateral, yield source, or bridged asset across at least nine major protocols. Each had to scramble independently:

Market-wide impact:

• Total DeFi TVL dropped ~$13B over 48 hours (CoinDesk; from ~$99.5B to ~$86.3B per DefiLlama snapshots).
• Aave’s TVL specifically fell $6–8.5B depending on snapshot window — Unchained reports $6.6B, CoinDesk reports $6B, blockchain.news measures $8.45B against a different baseline.
• AAVE token: -16% to -18%.

LayerZero’s official statement claims “no contagion” to other applications or cross-chain assets — a self-serving framing belied by the protocol-level pause list above. Other LayerZero-OFT-bridged assets did not get drained, but the bridge model’s confidence was meaningfully damaged across the network.

The most striking number is the Aave TVL outflow. Aave did everything procedurally correct — Guardian freeze initiated in 77 minutes, governance posts, transparent incident reporting — and still bled $6B+ in user deposits because the category of risk (third-party LRT collateral) was suddenly visible. This is the cost of being the lender of record for an asset whose safety you do not fully control.

7. The Kelp / LayerZero Configuration Dispute

The most important second-order story is the public disagreement over who chose the 1-of-1 DVN.

7.1 LayerZero’s Position

From LayerZero’s official statement:

“LayerZero and other external parties previously communicated best practices around DVN diversification to KelpDAO. Despite these recommendations, KelpDAO chose to utilize a 1/1 DVN configuration.”

LayerZero’s framing: the protocol functioned as designed; the failure was an application-layer configuration choice made over LayerZero’s documented advice. Forward commitment: LayerZero will no longer sign messages from any application running a 1-of-1 configuration, forcing network-wide migration.

7.2 Kelp DAO’s Position

Kelp’s response, per CoinDesk’s coverage:

• The 1-of-1 configuration is the default in LayerZero’s quickstart guide and default GitHub configuration.
• Kelp received no specific recommendation since July 2024 to change its single-verifier setup.
• Approximately 40% of LayerZero protocols currently use the same 1/1 structure — making it a de facto standard, not an unusual choice.

7.3 Independent Assessment

• Artem Krasnobaev (banteg): confirmed via direct review that LayerZero’s public deployment code uses single-source verification defaults across multiple chains.
• Zach Rynes (Chainlink): accused LayerZero of “deflecting responsibility” for what was, at root, a compromise of LayerZero Labs’ own infrastructure (the RPC nodes that were poisoned were the ones LayerZero’s DVN itself queried).

7.4 Reading the Dispute

Both narratives can be true simultaneously:

1. LayerZero did recommend multi-DVN setups in best-practice documentation.
2. LayerZero also ships defaults that produce 1-of-1 setups, and a large fraction of the network is deployed that way.
3. The actual compromise was at LayerZero Labs’ RPC layer (binary tampering on op-geth nodes) plus LayerZero’s failover logic accepting forged data when DDoS’d.

A useful frame: the configuration choice set the blast radius (1 verifier = total compromise), while LayerZero’s RPC infrastructure was the actual point of failure. Kelp’s choice was the “amplifier”; LayerZero’s RPC compromise was the “trigger.” Allocation of responsibility depends on whether one believes secure defaults are the platform’s job or the integrator’s job — a question the broader Ethereum ecosystem has been working out for years.

8. Recovery Options on the Table

As of April 20, no recovery or compensation plan has been formally announced. The community has surfaced four broad options:

Option A — Socialize losses across all rsETH holders. Mainnet and L2 rsETH each take a uniform ~15% haircut. Aave bad debt: $123.7M. Politically simplest; spreads the pain widely.

Option B — Concentrate losses on bridged rsETH (L2-only). Mainnet rsETH remains fully backed; L2 holders take a ~73.5% haircut on the affected portion. Aave bad debt: $230.1M, concentrated on Mantle, Arbitrum, Base, Ink. Strictly follows the “where was the money actually drained from” logic, but produces severe concentration on L2 holders who had no part in the configuration choice.

Option C — Snapshot restoration. Attempt to roll holder balances back to a pre-hack state and reissue. DefiLlama’s pseudonymous founder 0xngmi flagged this as “very hard to do” in practice — rsETH has moved through DEX pools, lending markets, and yield strategies in the interim.

Option D — Negotiated bounty. Some analysts have floated a 10–15% bounty to the attacker in exchange for the remainder. Politically toxic in light of the Lazarus attribution (paying a sanctioned actor would create severe legal exposure for any party touching the funds).

The longer the recovery decision sits, the more attacker positions on Aave drift toward forced liquidation as WETH utilization stays pinned at 100% — creating pressure to act before the bad debt crystallizes through liquidation rather than governance.

9. Accountability: A CROPS Audit

The Ethereum Foundation’s CROPS framework defines five properties Ethereum exists to protect:

• C — Censorship resistance
• R — Resistance to capture
• O — Open source
• P — Privacy
• S — Security

This incident is primarily a Resistance-to-Capture and Security failure, with secondary Open-Source implications. Censorship resistance and Privacy are not directly implicated (though Tornado Cash served the attacker’s pre-funding — a Privacy-vs-Security tension worth noting separately).

The audit below maps each party’s accountability against the relevant CROPS pillars and proposes mitigations.

9.1 Kelp DAO

9.2 LayerZero

9.3 Aave

9.4 The Ecosystem

The deeper lesson cuts across CROPS as a whole. Cross-chain LRTs concentrate value across many independent protocols whose individual security postures sum to far less than the headline TVL suggests. The CROPS framework treats Ethereum mainnet as the sanctuary; assets that claim to be Ethereum-equivalent on remote chains are only as sanctuary-grade as the bridge holding the backing.

The restaking narrative as amplifier

It is worth being precise about the restaking layer’s role. The exploit itself is not a restaking failure: EigenLayer slashing did not trigger, no AVS misbehaved, and a vanilla LST bridged through the same 1-of-1 LayerZero DVN would have been drained identically. The restaking primitive is technically blameless for the trigger.

But the LRT business model is not blameless for the blast radius. Three conditions made this contagion materially worse than a comparable LST drain would have been:

1. Cross-chain proliferation as product strategy. LRTs compete on “earn layered yield everywhere.” That pitch pushed Kelp to bridge rsETH to 20+ chains — far more aggressive deployment than vanilla LSTs typically pursue. Bridge surface area scales with chain count; more routes = more configuration choices that can be wrong.
2. Deep DeFi composability. rsETH was wired into Aave (11 deployments), Compound, Euler, Fluid, SparkLend, Lido Earn, Morpho, and Kamino — most as collateral, several as yield primitives. LRTs sit deeper in the composability stack than LSTs because the yield narrative demands re-use. The $6–8.5B Aave TVL outflow happened because the category of risk (LRT-as-collateral) suddenly became visible across that whole stack at once.
3. The “backed by mainnet ETH” illusion. LRT marketing frames the asset as ETH-equivalent, eliding the actual risk stack: validator slashing + EigenLayer/AVS slashing + operator delegation + bridge custody. L2 holders of rsETH were holding a bridge claim on a single Ethereum adapter contract — a fact most of them had not priced.

The deeper structural critique is that LRT issuance is concentrated in a small number of large protocols (Kelp, Renzo, EtherFi, Puffer), each of which becomes a single-point dependency for huge swaths of DeFi. It is the same anti-pattern as the bridge — centralization disguised as decentralization — and the CROPS Resistance-to-Capture failure runs through both layers simultaneously. The restaking narrative did not pull the trigger, but it built the room.

Ecosystem mitigations

1. Bridged-asset attestation standard. A public schema for bridges to publish (DVN config, verifier identities, rate limits, custody ratio) and for lending markets to consume programmatically.
2. Upgrade the bar for “blue-chip” status. A token’s blue-chip status today is largely market-cap and TVL based. It should also include bridge-level fault tolerance. An asset bridged via 1-of-1 is not blue-chip on the L2, regardless of its mainnet status.
3. Default-secure tooling. Cross-chain SDKs, wallet libraries, and bridge templates should ship with secure defaults. The maintainer owes the population-level outcome, not just the documentation footnote.
4. Sanctuary-grade collateral preference. Lending markets should prefer collateral whose backing lives on Ethereum L1 over collateral whose backing lives behind any bridge — the CROPS-aligned default. Bridged-collateral exposure should be priced (lower LTV, higher liquidation incentive) rather than treated as fungible.
5. LRT risk disclosure standard. LRT issuers should publish, per chain, the full risk stack: backing location, bridge configuration, AVS exposure, operator set, and slashing conditions. Lending markets should consume that disclosure machine-readably and price it into LTV and liquidation parameters. The “ETH-equivalent” framing should be retired.

10. Sources

Primary

• rsETH incident — 2026-04-18 — Aave Governance Forum
• rsETH Incident Report (April 20, 2026) — Aave Governance Forum
• KelpDAO Incident Statement — LayerZero

Coverage and Analysis

• 2026’s biggest crypto exploit: $292 million gets drained from Kelp DAO — CoinDesk
• Aave could face up to $230 million in losses after Kelp DAO bridge exploit triggers DeFi chaos — CoinDesk
• LayerZero blames Kelp’s setup for $290 million exploit — CoinDesk
• Kelp DAO claims LayerZero’s default settings caused the disaster — CoinDesk
• LayerZero says North Korea’s Lazarus likely behind Kelp DAO exploit — The Block
• DeFi losses top $600 million in weeks as Kelp DAO exploit drags TVL to one-year low — The Block
• The $293 million bug wasn’t in the code — WEEX
• LayerZero dispute deepens after $290 million rsETH bridge drain — Cryptonomist
• Kelp DAO $293M Exploit Triggers DeFi-Wide Contagion — Blockchain.news
• AAVE TVL plummets $6B after Kelp DAO hack — Crypto Briefing
• Kelp DAO Exploit Drains $292M in rsETH — FinanceFeeds
• DeFi Bleeds $7B In A Day After $290M Exploit — Stocktwits
• Kelp DAO rsETH Exploit Explained: Why AAVE Is Facing Heat — CoinGabbar
• Lazarus Group Suspected in $290M KelpDAO Hack — BanklessTimes

CROPS Framework

• Ethereum Foundation Unveils CROPS Principles — CryptoRank
• The Ethereum Foundation’s Cypherpunk Double-Down — Bankless
• Ethereum Foundation publishes mandate — The Block