A single forged cross-chain message drained 116,500 rsETH — $292 million — from Kelp DAO's LayerZero-powered bridge on April 18, 2026, at 17:35 UTC. The exploit, attributed to North Korea's Lazarus Group by both Chainalysis and LayerZero, is the largest DeFi theft of 2026. No smart contract vulne...
"DeFi United has secured sufficient ETH commitments to restore full backing." — Stani Kulechov, Founder, Aave
A single forged cross-chain message drained 116,500 rsETH — $292 million — from Kelp DAO's LayerZero-powered bridge on April 18, 2026, at 17:35 UTC. The exploit, attributed to North Korea's Lazarus Group by both Chainalysis and LayerZero, is the largest DeFi theft of 2026. No smart contract vulnerability was involved. The attacker compromised off-chain RPC infrastructure and exploited a 1-of-1 Decentralized Verifier Network (DVN) configuration — a single point of failure that required zero fault tolerance to bypass.
Within 48 hours, contagion effects erased $13.2 billion in total value locked across 20+ protocols. Aave absorbed between $177 million and $230 million in bad debt from attacker-deposited collateral. A coalition called DeFi United, led by Aave and backed by 14+ contributors, has since pledged 132,650 ETH (~$303 million) to restore rsETH backing. Arbitrum's Security Council froze $71 million in attacker funds. The incident exposes a structural weakness not in code, but in the configuration defaults and verification architecture underpinning cross-chain restaking infrastructure.
At 17:35 UTC on April 18, an attacker linked to the Lazarus Group's TraderTraitor sub-unit executed a multi-stage infrastructure attack against Kelp DAO's cross-chain bridge.
Phase 1: RPC Compromise. The attacker identified and compromised two internal RPC nodes operated by LayerZero — hosted on separate clusters — replacing the software running on them with malicious code. According to Chainalysis's forensic analysis, these nodes were among the endpoints queried by LayerZero's DVN when validating cross-chain messages.
Phase 2: DDoS Isolation. A simultaneous distributed denial-of-service attack targeted external RPC nodes the DVN relied on for redundancy. With those nodes knocked offline, the DVN failed over to the only endpoints it could reach: the two compromised internal nodes now controlled by the attacker.
Phase 3: Message Forgery. The attacker submitted a fabricated cross-chain message claiming a valid rsETH burn had occurred on a source chain. The 1-of-1 DVN configuration meant no second verifier was required to corroborate the message. Kelp's Ethereum bridge contract released 116,500 rsETH — approximately 18% of the token's circulating supply — to an attacker-controlled address.
Phase 4: Emergency Response. Kelp's emergency pauser multisig froze core contracts at 18:21 UTC, 46 minutes after the initial drain. Two subsequent drain attempts at 18:26 and 18:28 UTC, each targeting an additional 40,000 rsETH (~$200 million combined), reverted against the paused contracts. The multisig's response prevented a potential total loss exceeding $490 million.
The stolen rsETH was subsequently distributed across 20+ chains. On-chain tracking by Unchained shows $175 million was routed through THORChain into new wallet addresses. Arbitrum's Security Council froze 30,766 ETH ($71 million) that remained on Arbitrum One.
The exploit produced a contagion ratio of approximately 45:1. For every dollar stolen, $45 of additional capital exited the DeFi sector within 48 hours, according to CoinDesk reporting on DeFiLlama data.
Aave saw $8.45 billion in deposit withdrawals over the same period — far exceeding the direct exploit amount. The withdrawals reflected a generalized loss of confidence in liquid restaking token (LRT) collateral across lending markets, not merely rsETH-specific risk.
stETH, which was not directly affected by the exploit, traded below its ETH peg in the immediate aftermath as liquidity conditions tightened across Aave's ETH markets. The rsETH token itself depegged from its nominal exchange ratio of 1.07 ETH, though the precise discount varied across trading venues.
The $13.2 billion TVL decline affected protocols with no direct exposure to Kelp DAO or rsETH. This pattern mirrors prior DeFi contagion events — notably the $3.6 billion withdrawn from Curve Finance pools during the July 2023 Vyper exploit — but at significantly larger scale.
The attacker deposited 89,567 rsETH into Aave as collateral and borrowed approximately $190 million in ETH and related assets across Aave Ethereum Core and Aave Arbitrum. This created between $177 million and $230 million in bad debt, depending on recovery assumptions.
Aave's incident report, published April 20, identified eight affected positions across two deployments. The protocol froze rsETH markets and paused new deposits immediately. The bad debt figure represents the gap between the attacker's outstanding borrows and the now-devalued rsETH collateral backing them.
To clear the positions, DeFi United's technical proposal calls for temporarily lowering the rsETH oracle price via governance vote to enable controlled liquidations. This would allow the protocol to seize and redistribute the attacker's rsETH collateral at a discount, reducing the total bad debt requiring external funding.
Compound also reported exposure through rsETH collateral positions, though at a smaller scale. Compound DAO has proposed allocating up to 3,000 ETH to the recovery effort.
DeFi United is a coalition of seven core protocols — Aave, Lido, EtherFi, Ethena, Mantle, Ink Foundation, and BGD Labs — formed within 48 hours of the exploit. As of April 28, the coalition reported 132,650 ETH in capital and credit commitments, valued at approximately $303 million. The structure resembles an ad hoc lender-of-last-resort mechanism, with no single entity bearing the full cost.
| Contributor | Pledge (ETH) | Approx. USD | Status | |---|---|---|---| | Consensys / Joe Lubin | 30,000 | $69M | Pledged | | Mantle (credit facility) | 30,000 | $69M | Proposed | | Aave DAO (treasury) | 25,000 | $57M | Governance vote | | LayerZero | 10,000 | $23M | Committed | | EtherFi | 5,000 | $11.5M | Under discussion | | Stani Kulechov (personal) | 5,000 | $11.5M | Pledged | | Lido | 2,500 stETH | $5.8M | Proposed | | Compound DAO | 3,000 | $6.9M | Proposed | | Kelp DAO | 2,000 | $4.6M | Committed | | Renzo | — | $10M+ | Treasury deployed | | Babylon Foundation | — | $3M (USDT) | Planned | | Emilio Frangella (personal) | 500 | $1.15M | Pledged |
The restoration plan calls for converting committed ETH into rsETH in controlled tranches to limit market disruption, then transferring the rsETH to the affected lockbox contract. This would allow the bridge to resume operations once rsETH returns to its nominal 1.07 ETH exchange ratio.
By April 26, Arkham Intelligence reported approximately $160 million — roughly 80% of the minimum $200 million Aave estimated it needs — had been raised.
The exploit has triggered a public dispute between Kelp DAO and LayerZero over responsibility.
LayerZero's position: In an April 20 statement, LayerZero attributed the exploit to Kelp DAO's choice of a 1-of-1 DVN configuration and stated it had recommended multi-DVN setups. LayerZero emphasized that its messaging protocol itself was not compromised — only the verification layer configured by the application.
Kelp DAO's position: Kelp responded the same day, stating the 1-of-1 configuration was the default setting shipped for new deployments at the time of its Layer 2 expansion. Kelp characterized LayerZero's multi-DVN recommendation as post-hoc and argued the default should have been secure by design.
The DVN architecture delegates security decisions to the application layer: each protocol integrating with LayerZero independently chooses how many DVN nodes must simultaneously confirm a cross-chain message. The standard industry recommendation is at least a 2-of-3 configuration, requiring an attacker to compromise two independent nodes to forge a message. Higher-security setups (e.g., 5-of-9) can achieve 55% fault tolerance.
According to KuCoin's analysis, $293 million was lost without any code vulnerability — the exploit exposed a configuration blind spot that standard smart contract audits do not cover. Blockaid's technical writeup confirmed that auditing DVN configuration parameters was not part of any published audit scope for Kelp DAO's bridge deployment.
The Arbitrum Security Council's emergency freeze of 30,766 ETH ($71 million) on Arbitrum One, executed April 21, recovered approximately 25% of the stolen funds. Aave Labs and Kelp DAO subsequently submitted a joint proposal requesting Arbitrum DAO to release the frozen funds to the DeFi United recovery pool.
The freeze has reignited debate over Layer 2 governance centralization. The Arbitrum Security Council — a 12-member multisig with 9-of-12 execution authority — exercised emergency powers designed for protocol-level threats, not fund recovery. Critics argue the action sets a precedent for chain-level asset seizure that conflicts with censorship-resistance guarantees.
Proponents counter that without the freeze, the full $292 million would have been laundered through mixing protocols, as demonstrated by the $175 million already routed through THORChain. The tension between immutability principles and practical loss recovery remains unresolved.
The EigenLayer-dominated restaking sector held $16.26 billion in TVL at the time of the exploit, with EigenLayer commanding $15.26 billion (93.9% market share) and 4.36 million ETH. The Kelp exploit did not affect EigenLayer's core contracts, but it exposed systemic risk in how restaked assets are bridged across chains.
Bridge security as the weakest link. The restaking value chain — stake ETH, receive a liquid restaking token, bridge that LRT to another chain for yield — introduces bridge risk at the final step. The security of the underlying staking mechanism is irrelevant if the bridge can be compromised to mint unbacked tokens.
Configuration risk vs. code risk. Protocols routinely undergo smart contract audits before deployment. DVN configuration parameters, RPC endpoint security, and failover logic — the exact vectors exploited in this attack — are typically outside audit scope. This gap affects every protocol using LayerZero's messaging infrastructure.
Composability amplifies contagion. The 45:1 contagion ratio demonstrates how DeFi's composable architecture — where LRTs serve as collateral in lending markets, which in turn back stablecoins and structured products — creates multiplicative loss potential from a single exploit.
The Ethereum Foundation completed its 70,000 ETH staking target ($143 million) in early April 2026, and institutional staking continues to grow. EtherFi's $3 billion validator liquidity agreement with ETHGas, signed April 15, underscores institutional appetite. But the Kelp exploit demonstrates that bridged restaking assets carry infrastructure risk that current audit frameworks do not capture.
$292 million stolen, zero code vulnerabilities exploited. The attack targeted off-chain RPC infrastructure and a 1-of-1 DVN configuration, not smart contracts. Standard DeFi audits would not have flagged the vulnerability.
45:1 contagion ratio. $13.2 billion in TVL exited DeFi within 48 hours — 45x the direct exploit amount — demonstrating how composable collateral chains amplify single-point failures.
$303 million recovery coalition formed in 10 days. DeFi United, comprising 14+ contributors led by Aave, assembled the largest coordinated DeFi bailout to date. Approximately 80% of the minimum recovery target was raised by April 26.
$71 million frozen by Arbitrum Security Council. The emergency asset freeze recovered ~25% of stolen funds but raised unresolved questions about Layer 2 governance centralization and censorship resistance.
Default configurations shipped insecure. The dispute between Kelp DAO and LayerZero over who bears responsibility for the 1-of-1 DVN default highlights an industry-wide gap in secure-by-default design for cross-chain infrastructure.
Audit scope insufficient for infrastructure risk. No published audit of Kelp's bridge deployment covered DVN configuration, RPC endpoint security, or failover behavior. The restaking sector's $16.3 billion TVL rests on verification assumptions that remain largely unaudited.
The Kelp DAO exploit is a stress test that the restaking sector did not seek but cannot afford to ignore. The $292 million loss itself is containable — DeFi United's $303 million commitment suggests the ecosystem can absorb the direct damage. The more consequential finding is structural: $16.3 billion in restaked assets traverses cross-chain infrastructure whose security depends on configuration decisions that sit outside the scope of standard audits, are often set to insecure defaults, and are verified by a single entity.
The 45:1 contagion ratio confirms that in a composable financial system, the weakest infrastructure component determines the risk profile of the entire stack. For institutional capital entering restaking through vehicles like EtherFi's $3 billion ETHGas agreement, the question is no longer whether the staking mechanism is secure, but whether every link in the cross-chain value chain — from DVN configurations to RPC failover logic — meets the same standard.
DeFi United's rapid formation offers a precedent: the sector can coordinate emergency responses at scale. Whether it can prevent the next exploit by addressing infrastructure configuration risk before deployment — rather than assembling bailout coalitions after — remains the open question.