Cross-chain bridges have hemorrhaged $340.7 million across 14 major exploits in 2026 through mid-August, according to data compiled by PeckShield and CoinGabbar. The figure accounts for roughly 44% of all DeFi-related losses during the period, despite bridges holding a fraction of total DeFi TVL....
"We've done a terrible job on comms over the past three weeks." — LayerZero Labs, public apology following the $292M Kelp DAO bridge exploit (May 8, 2026)
Cross-chain bridges have hemorrhaged $340.7 million across 14 major exploits in 2026 through mid-August, according to data compiled by PeckShield and CoinGabbar. The figure accounts for roughly 44% of all DeFi-related losses during the period, despite bridges holding a fraction of total DeFi TVL. The attack surface has shifted: exploits no longer target smart contract code. They target signing keys, RPC nodes, verifier configurations, and deposit validation logic — infrastructure layers that standard code audits do not cover.
The single largest incident — the $292 million Kelp DAO rsETH exploit on April 18 — exposed a systemic design flaw in LayerZero's cross-chain messaging stack. A Dune Analytics review found that 47% of LayerZero's ~2,665 active OApp contracts operated with the same 1-of-1 verifier configuration that enabled the attack, representing a combined exposure exceeding $4.5 billion. The pattern repeated in smaller incidents throughout the summer: THORChain ($10.8M in May), AFX Trade ($24.2M in July), Verus Protocol ($7.5M, exploited twice), and the Coreum-XRPL bridge ($200K XRP in August).
Bridge TVL dropped below $45 billion in late June, down from peaks earlier in the year, as capital migrated toward intent-based architectures (Across, deBridge) that minimize custodial risk by design.
PeckShield tracked eight major cross-chain bridge attacks through mid-May, totaling $328.6 million in losses. Additional incidents in June, July, and August pushed the cumulative total to $340.7 million across 14 exploits, according to CoinGabbar data through early August.
| Date | Protocol | Loss | Attack Vector | |------|----------|------|---------------| | Feb 2026 | IoTeX ioTube | $4.4M | Private key compromise | | Feb 2026 | CrossCurve | ~$3.0M | Validation gaps | | Feb 2026 | Hyperbridge | $2.5M | Exploit (unspecified) | | Apr 18 | Kelp DAO (rsETH/LayerZero) | $292M | RPC poisoning + DDoS of single DVN | | May 15 | THORChain | $10.8M | TSS key extraction via malicious validator | | May 2026 | Verus-Ethereum (1st) | ~$4.0M | Reserve verification bypass | | Jul 22 | AFX Trade (Arbitrum) | $24.2M | 5-of-N validator key compromise | | Jul 22 | Verus-Ethereum (2nd) | $7.5M | Same bug re-exploited | | Aug 9 | Coreum-XRPL | ~$200K (XRP) | Fake deposit label spoofing |
Bridge exploits accounted for 42% of May 2026's ~$70 million total crypto exploit losses, a disproportionate share given bridges' fraction of total DeFi TVL. July losses reached $97 million across all crypto exploit categories, with bridge attacks contributing over $35 million, per PANews data.
On April 18, an attacker extracted 116,500 rsETH (worth ~$292 million) from Kelp DAO's cross-chain bridge, which used LayerZero's V2 messaging infrastructure. The attack did not exploit a single line of smart contract code.
Attack sequence, per Chainalysis and LayerZero post-mortems:
LayerZero initially attributed the attack to North Korea's Lazarus Group (specifically its TraderTraitor subunit), according to a post-mortem published April 20.
Three weeks of public dispute followed. Kelp DAO claimed LayerZero approved the 1-of-1 verifier configuration during integration. LayerZero maintained its documentation recommended multi-verifier setups. On May 9, LayerZero reversed course, stating it "made a mistake" by allowing its verifier infrastructure to secure high-value assets in this configuration. The company pledged $23 million to a "DeFi United" recovery fund.
DeFi TVL dropped more than $13 billion in two days following the attack, according to CoinDesk reporting.
A Dune Analytics review of ~2,665 active LayerZero OApp contracts — published after the Kelp incident — revealed the structural depth of the problem:
The 1-of-1 setup means a single compromised verifier node can authorize fraudulent cross-chain messages. This is the exact vector used in the Kelp attack. The Dune data suggests the Kelp incident was not an anomaly of poor configuration but a reflection of the default security posture across nearly half the LayerZero ecosystem.
According to The Defiant, LayerZero's default DVN setup during integration was the 1-of-1 configuration, which developers adopted without upgrading. The Techmeme summary noted that in April, approximately 47% of LayerZero OApps had the same default setup.
On May 15, an attacker drained $10.8 million from THORChain's Asgard vaults across Bitcoin, Ethereum, BNB Chain, Base, Avalanche, Dogecoin, Litecoin, Bitcoin Cash, and XRP — nine chains in total, according to TRM Labs.
Attack mechanism:
The attacker entered THORChain's validator set as a newly churned node operator two days prior to the exploit. During legitimate signing ceremonies, the malicious node extracted cryptographic key fragments from THORChain's GG20 Threshold Signature Scheme (TSS). Over multiple signing rounds, the attacker reconstructed the full private key of an Asgard vault and signed unauthorized outbound transactions.
THORChain's automatic solvency detection systems triggered within minutes, halting signing and trading without human intervention. Node operators then coordinated through Discord, deploying manual pauses and Mimir governance votes to bring the network to a controlled halt within approximately two hours, according to THORChain's official exploit report.
On-chain investigator ZachXBT and PeckShield flagged the attack before THORChain's public acknowledgment. RUNE dropped 15% on the news, per CoinDesk reporting.
The THORChain exploit illustrates a different failure mode from Kelp: here, the cryptographic primitives themselves were exploited through a protocol-level vulnerability in the TSS implementation, rather than infrastructure compromise.
On July 22, two bridges were exploited within seven hours, resulting in combined losses of $31.6 million, according to CoinInsider and TradingView reporting.
AFX Trade ($24.2M): At 21:30 UTC, attackers withdrew 24,150,000 USDC from AFX Trade's custodial bridge on Arbitrum. According to Blockaid, the attackers had obtained private keys from five of the bridge's validators, achieving the quorum necessary to authorize the withdrawal. The attack vector was social engineering and key management failure, not code exploitation.
Verus Protocol ($7.5M): A hacker reused the same vulnerability from a May exploit to drain $7.54 million from the Verus-Ethereum bridge. The bridge approved eight withdrawals without verifying that matching reserves backed them — the identical class of bug that caused the earlier Verus incident. The protocol had not patched the reserve verification logic in the intervening two months.
PANews reported July 2026 total crypto security losses at approximately $97 million, with concentrated cross-chain bridge attacks accounting for over $35 million.
On August 9, an attacker drained 199,916 XRP from the Coreum-XRPL bridge in 97 minutes through 94 consecutive fraudulent transactions, according to CryptoRank and CCN reporting.
Attack mechanism:
The attacker never sent real XRP to the bridge. Instead, they transferred the bridge's own token between two self-controlled wallets, attaching a fake deposit label. The bridge's deposit verification software parsed the transaction history, incorrectly interpreted the labeled transfer as a legitimate incoming deposit, and credited the attacker's account. Withdrawals then proceeded through the normal multisig process — 17-of-28 relayer keys signed each outgoing payment.
The bridge was left with 493.5 XRP, down from ~200,000 XRP. XRP fell to $1 for the first time since November 2024 following the disclosure, per Benzinga. The XRP Ledger itself was not compromised; the vulnerability resided entirely in the bridge's off-chain validation logic.
| Year | Total Crypto Hack Losses | Bridge-Specific Losses | Notable Incidents | |------|-------------------------|----------------------|-------------------| | 2022 | ~$3.8B | ~$2.0B | Ronin ($624M), Wormhole ($320M), Nomad ($190M) | | 2023 | ~$1.7B | Declining | Multichain ($126M) | | 2024 | ~$2.2B | ~$300M | Category share decreased | | 2025 | ~$3.4B | Not dominant | Bybit exchange hack dominated ($1.5B) | | 2026 (YTD) | ~$972M (H1) | $340.7M (14 exploits) | Kelp DAO ($292M), THORChain ($10.8M) |
Sources: Chainalysis, PeckShield, CoinGabbar, CoinDesk.
Bridge-specific losses peaked at ~$2 billion in 2022 and declined to ~$300 million by 2024 as lock-and-mint bridges fell out of favor. The 2026 figure of $340.7 million marks a reversal, driven almost entirely by the Kelp DAO incident. Excluding Kelp, bridge losses in 2026 total approximately $48.7 million — roughly in line with the post-2022 trend.
The broader pattern: the total number of crypto security incidents is accelerating (207 incidents in H1 2026 vs. 83 in H1 2025), but dollar losses per incident are declining as high-value targets adopt better security practices. Bridges remain the exception.
The recurring theme across 2026's bridge exploits is the failure of custodial trust models — bridges that hold locked assets and rely on validator sets, multisigs, or single verifiers to authorize releases. The attack surface maps directly to the number and quality of signing keys, RPC nodes, and verification layers.
A structural alternative has gained traction: intent-based bridges. Protocols like Across and deBridge operate with near-zero TVL because liquidity providers front funds to users without the bridge ever taking custody. The user states an "intent" (e.g., move 10 ETH from Arbitrum to Ethereum), and a solver fulfills it from their own balance, settling later through an optimistic verification process.
According to DeFiLlama, bridge TVL dropped below $45 billion in late June, down from peaks earlier in the year. The decline reflects both post-exploit capital flight and migration toward intent-based designs that minimize custodial exposure.
The trade-off: intent-based bridges depend on solver liquidity and are slower for large transactions. They are not universally applicable. But the security model is structurally different — there is no pool of locked assets for an attacker to drain.
An independent vulnerability discovered by security firm Hexens and reported July 4 underscored the broader infrastructure risk: researchers found a flaw in the Aptos blockchain that could have put up to $70 billion in cross-chain assets at systemic risk. They simulated the attack with over 90% success rate using a $3,000 server. The vulnerability was patched within hours of disclosure, with no funds lost, according to CoinDesk.
The 2026 bridge exploit data tells a specific story: the attack surface has migrated from code to infrastructure. Smart contract audits — the industry's primary security mechanism — do not cover RPC node integrity, validator key management, TSS implementations, or deposit verification logic. Of the five largest bridge incidents this year, none involved a smart contract vulnerability.
The economic incentive structure is clear. Bridges aggregate cross-chain liquidity into concentrated pools protected by a small number of signing keys or verification nodes. An attacker who compromises the verification layer — whether through RPC poisoning, social engineering, or protocol-level TSS exploitation — gains access to the entire pool. The expected value of attacking a bridge with $200 million in TVL and a 3-of-5 multisig is materially higher than attacking a lending protocol of equivalent size with on-chain liquidation mechanisms.
The Dune data showing 47% of LayerZero OApps on 1-of-1 verifier configurations suggests the industry's default security posture is calibrated for convenience, not adversarial conditions. LayerZero's admission that this was its default setup — and its subsequent $23 million remediation fund — acknowledges the gap between documented best practices and actual deployment patterns.
Intent-based architectures offer a structural alternative by eliminating the custodial pool, but they introduce new dependencies on solver networks and settlement layers. The transition is underway but incomplete.
The data implies that bridge security is not primarily a technology problem. It is a configuration, key management, and infrastructure monitoring problem. Until the industry's security apparatus extends beyond code audits to cover these layers, bridges will remain disproportionately vulnerable relative to their share of DeFi TVL.