Verus-Ethereum Bridge
Why it happened
Attack class: Bridge exploitThe bridge accepted a signed but unbacked import: the Ethereum contract and Verus disagreed about the same data, and nothing checked that the payout matched value locked on Verus.
What would have stopped it
- Reconcile every payout against value locked on the source chain.
- Use one canonical encoding that both chains must parse identically.
Not scored before the exploit
BlackHart had not scored Verus-Ethereum Bridge before this exploit.
An attacker drained about $7.54 million from the Verus-Ethereum bridge, the second time the same contract was emptied in 66 days. They got the bridge's Ethereum contract to pay out ETH, tBTC, DAI, USDC, USDT, scrvUSD, EURC and MKR for a transfer that never happened on Verus, even though every signature checked out. Bridge reserves on Ethereum were hit, which left Verus-side bridged ETH and tBTC only partly backed; Verus has said it will repay holders with restitution tokens. The attacker swapped everything to about 3,916 ETH and put it all through Tornado Cash within two hours.
A fresh wallet sent an import to the Verus-Ethereum bridge that released about $7.54 million in one transaction to a receiving wallet, which swapped it to ETH and deposited about 3,916 ETH into Tornado Cash.
Attack flow
All 4 on-chain stepsThe attack in 3 steps. Step 1, Unbacked import: Verus-Ethereum bridge (bridge). Step 2, Swap: Receiving wallet (hub), about $7.54M reached this step. Step 3, Mixer: Tornado Cash (exit).
Step 1: Unbacked import
- BridgeVerus-Ethereum bridge1,137 ETH, 71.5 tBTC, stablecoins, MKR~ $7.54M in one transaction0x71518580...cD7F63
Step 2: Swap
- HubReceiving walletswaps via Relayabout 3,916 ETH0xcfd0a207...142d54
Step 3: Mixer
- ExitTornado Cashabout 3,916.1 ETH04:45 to 05:30 UTC0xd90e2f92...24F31b
Why each step worked
- Ethereum side read signed data differently from Verusvalid notary signatures, inconsistent interpretation
- No check that payout matched value locked on Verusmissing controlsame class of flaw as the May drain
- No holding period on large payoutsmissing controlrelease settled instantly
- Bridge refilled before an independent reviewmissing controlrecovered funds returned two weeks earlier
Full forensic detail
Step-by-step reconstruction, root cause, counterfactuals, remediation, and disclosure timeline.
Exploit anatomy
Root cause
The Verus-Ethereum bridge releases Ethereum reserves when it receives an import from Verus that is backed by a notarized summary of the Verus ledger. The checks it ran, notary signatures and Merkle proofs, all passed, per Blockaid, CertiK and Halborn. What failed was consistency between the two sides: the Ethereum contract could read the signed data differently from Verus nodes, and it did not confirm that the value it paid out matched value actually locked or exported on Verus. That let a transfer that Verus never made be honored on Ethereum. No key was stolen and no cryptography was broken, per Verus. Blockaid tied the attack to the same contract, entry path and class of flaw as the May 2026 drain of about $11.6 million, which means the earlier fix did not close the underlying gap; Verus later said it was not quite the same bug. This repeat came 66 days after the first, after Verus had moved recovered funds back into the bridge. As of the latest public updates Verus had not reopened the bridge, its planned Ethereum contract upgrade was still pending, and no independent audit had been commissioned, so the details of the flaw are deliberately kept general here.
Prevention analysis
Valid signatures only prove that data was signed, not that it moves real value. A check that each release matches value exported from Verus would have rejected an import that Verus never sent.
The loss came from the Ethereum contract and Verus nodes disagreeing about the same signed data. Rejecting anything ambiguous on both sides removes that whole class of attack.
The drain settled in one transaction. Verus's own planned fix, an 8 to 12 hour holding period in which witnesses can halt a transfer, would have given time to stop it.
Recovered funds went back into the same contract two weeks before the second drain. An outside review of the May fix, before restoring reserves, could have caught that the gap was still open.
Similar incidents
The same bridge contract lost about $11.6 million in May to an import the Ethereum side read differently from Verus. Same class of flaw, different attacker, and the fix in between did not hold.
Syscoin's relay read a burn proof differently from Syscoin Core and released 5 billion SYS that were never burned. Same root: two layers of one bridge disagreeing about what signed data says.
Nomad accepted messages that were never sent and paid them out from reserves. Same outcome: the destination chain released funds without proof that the source chain locked them.
Remediation
Timeline
Get your protocol scored across 12 dimensions, or request ongoing coverage.