DeFi Bridge Exploits: Inside the $840M 2026 Hack Crisis

Three of the four largest DeFi hacks in 2026 did not involve a single line of flawed Solidity. The smart contracts executed exactly as written. Audits had signed off on the code. The attackers targeted something else: the bridges connecting blockchains, the administrative keys controlling those bridges, and the people who held those keys.

Through April 2026, more than $840M had been stolen across 47 separate incidents — making April the most-hacked month in crypto history by incident count. Kelp DAO lost $293M through forged cross-chain messages. Drift Protocol lost $285M to a six-month social engineering campaign run by North Korean state hackers. Step Finance lost $27M after a phishing attack compromised an executive's device. In each case, the attack surface was not protocol logic — it was infrastructure and people.

Why Bridges Hold the Biggest Targets in DeFi

Cross-chain bridges hold large pools of assets by design. Bridge TVL hit $21.94 billion as of March 2026, making individual bridges some of the largest single custody concentrations in DeFi. That concentration is the point — bridges need liquidity to process transfers — but it also makes them obvious targets.

Every bridge faces the same structural problem: it has no direct visibility into what happened on the originating chain. To know that a user locked 10 ETH on Ethereum and therefore deserves 10 ETH-equivalent on Arbitrum, the bridge must trust some external mechanism — a decentralized verifier network (DVN), a set of oracles, or a relay — to confirm the cross-chain message is legitimate. The security of that verification layer, not the contract code itself, determines whether a bridge can be deceived.

Most users never see this layer. A bridge interface looks identical whether it runs one verifier or five independent ones. Audit reports typically cover smart contract logic without testing whether the DVN or oracle configuration resists spoofing. Two bridges can both pass a code audit while carrying very different risk profiles based on nothing visible in the interface.

How Bridge Message Forgery Drained $293M from Kelp DAO

In April 2026, Kelp DAO's liquid restaking token rsETH suffered the largest DeFi exploit of the year. The protocol used LayerZero, a cross-chain messaging framework, to receive instructions across chains. The vulnerability was not in LayerZero's code — it was in how Kelp had configured its verification setup.

Kelp's LayerZero integration used a single DVN: only one verifier was required to validate each incoming cross-chain message. One verifier is one point of failure. The attacker found a method to compromise or spoof that single verifier and forge what appeared to be a legitimate cross-chain instruction. That message told the bridge to release funds. The contract executed exactly what it was designed to do.

The DVN count directly controls the cost of message forgery:

Configuration	Required Approvals	Minimum Attack Requirement
Single DVN	1 of 1	Compromise 1 verifier
2-of-3 DVN	2 of 3	Compromise 2 independent verifiers
3-of-5 DVN	3 of 5	Compromise 3 independent verifiers

Multi-DVN configurations don't eliminate bridge risk but raise the attack cost to a level where most exploits become unprofitable. Kelp's single-DVN setup gave the attacker a $293M payday from compromising one component.

The aftermath showed how bridge risk propagates. Aave had accepted rsETH as collateral. After rsETH's value collapsed in the wake of the exploit, Aave faced up to $230M in potential losses from borrowers whose rsETH-backed positions were now underwater. A bridge exploit can cascade through every protocol that accepts the bridge's output token as collateral, even if those downstream protocols have no code vulnerabilities at all.

This composable exposure is easy to miss. A user who never bridged anything, who deposited into a well-audited lending market, can find themselves exposed to a bridge attack through collateral they never directly held.

Social Engineering and Key Compromise: The Drift Protocol Attack

The second-largest DeFi hack of 2026 had nothing to do with bridges. Drift Protocol, a Solana-based decentralized exchange, lost $285M after North Korean state hackers spent six months systematically obtaining access to its administrative keys.

TRM Labs attributed the attack to the Lazarus Group. The attackers did not look for contract vulnerabilities. They attacked Drift's organization.

The campaign ran for approximately 24 weeks before the exploit:

• Hackers posed as legitimate trading firms and institutional counterparties
• They established communication with protocol contributors through standard industry channels
• They deposited substantial capital into Drift to appear as high-value users building credible trading history
• They attended industry conferences, building in-person trust with team members over months
• They eventually obtained access to an administrative key carrying protocol upgrade permissions

Once they held the key, the actual exploit took minutes. The six-month trust-building campaign was the attack — the on-chain transaction was the final step in a process that was almost entirely off-chain.

Drift's smart contracts were sound. No auditor reviewing the code would have found this vulnerability, because the vulnerability was not in the code.

Step Finance's $27M loss in the same period followed a shorter version of the same logic. A phishing campaign compromised an executive's device and gave attackers access to treasury accounts. Again, no contract flaw was involved. The threat model was not "can we exploit this code?" — it was "can we obtain legitimate key access from someone who already has it?"

Looking at the four largest exploits of 2026:

• Kelp DAO, $293M: Bridge message forgery via single-DVN misconfiguration
• Drift Protocol, $285M: Six-month social engineering campaign, state-sponsored
• Step Finance, $27M: Phishing compromise of an executive's device
• Verus Bridge, $11.58M: Cross-chain bridge drain, attack continued after discovery

Three of four bypassed contract code entirely.

Reading Protocol Risk Before You Bridge

For users moving assets across chains, the security question is no longer "has this been audited?" It is "what is the verification architecture, and who controls the keys?"

DVN configuration is public information for LayerZero-based bridges. LayerZero's documentation and scan tools expose which verifiers a given integration requires. A bridge requiring a single DVN carries meaningfully higher message-forgery risk than one requiring two or three independent verifiers. This is checkable before depositing — it does not require reading code.

Beyond DVN setup, the signals worth checking before using any protocol with significant TVL:

• Key management: Are upgrade and admin keys held by a multisig? What is the signing threshold? A 2-of-3 multisig where all three signers share an office provides weaker protection than geographically and organizationally separated key holders.
• Upgrade timelock: A protocol where admin keys can push immediate upgrades with no delay removes the last human check before an attacker acts on a compromised key. Even a 48-hour timelock gives users, white-hat researchers, and governance participants a response window.
• Audit scope: Many audits cover smart contract logic but not the integration layer — the DVN configuration, oracle setup, or admin key structure. Asking what was explicitly in scope is more informative than checking the name of the audit firm.
• Collateral exposure: If you hold assets in a protocol that accepts bridged tokens as collateral, your risk includes the security of the bridge that minted those tokens. Checking what collateral types a lending market accepts tells you which bridge architectures you are indirectly exposed to.

The Kelp DAO and Drift Protocol cases are not outliers. They represent a structural shift in how DeFi is being attacked. Code audits, formal verification, and bug bounties address contract risk. They do not address bridge message forgery, compromised admin keys, or social engineering against key holders. For any bridge or protocol holding substantial TVL, those are now the primary attack surfaces.

The clipper malware post covers the user-side version of this same pattern — device-level attacks that give attackers wallet access without exploiting any protocol. Both social engineering against protocol teams and device compromise against individual users bypass code by going around it. Spending five minutes reviewing DVN setup and key management before bridging significant assets delivers more risk reduction than reading another audit report.