Base DeFi vault drained of $6M in wstETH after multisig whitelist takeover

On October 4, 2026, an unnamed DeFi vault on the Base network lost roughly $6 million in wrapped staked Ether after someone manipulated the vault's own administrative controls to grant a malicious contract trusted status. According to a report on dev.to, the theft took less than 20 minutes from the final configuration change to the last withdrawal, and it did not depend on any flaw in the vault's financial logic.

How the attack unfolded

The targeted application was an unidentified vault holding a position on Aave's V3 market on Base, an Ethereum layer-2 network built on the OP Stack. Its assets were aBaswstETH tokens, the interest-bearing receipts Aave issues when Lido's wrapped staked Ether (wstETH) is supplied to the lending pool. The vault was built as an OpenZeppelin proxy and governed by a Safe multisignature wallet in a 3-of-7 configuration.

Per the timeline reconstructed from on-chain monitoring data cited by dev.to, the multisig executed a transaction at 08:52 UTC that removed a newly deployed, previously unseen smart contract from the vault's lending whitelist. One minute later, at 08:53 UTC, the same contract was added back. Whether that back-and-forth came from a compromised signing device, a deceived signer, or deliberate internal action remains officially unconfirmed.

The effect was immediate: once listed, the contract could interact with the vault as an approved counterparty. Within roughly 19 minutes it pulled 1,783.067 aBaswstETH out of the vault across six separate transfers, then redeemed those receipts directly through Aave V3 for approximately 1,783 wstETH.

Detection and tracing

Security firms responded quickly. Blockaid initially flagged the anomaly and put losses at around $2.02 million spanning roughly four transactions; as the exploit continued and more transfers were identified, the estimate climbed past $6 million. Spot On Chain and PeckShield each independently traced the stolen assets to a suspected attacker address on Base, reported as 0x0B5126…B034.

A permissions problem, not a protocol bug

Analysis by security researchers found no evidence that Aave's core lending contracts or the Base network itself were breached, according to dev.to. The failure sat entirely in the application-level authorization controls of this particular vault. Whitelists exist to restrict a protocol's interactions to pre-approved, audited contracts; by coaxing the multisig into approving a hostile contract, the attacker sidestepped the vault's protective mechanisms without touching the lending logic underneath.

Exactly how the multisig was compromised is still unknown. The seven signers have not been publicly identified, and no official post-mortem has established whether the root cause was a compromised administrative key, a misconfiguration in the access-control function, or a deeper issue in the proxy's upgrade logic. Whatever the entry point, the outcome was a failure of permission management.

Contained losses, real ripples

A $6 million hit is painful for the vault's users but modest against the backdrop of 2026's larger crypto breaches; dev.to points to the $387.5 million Bitget hack weeks earlier for comparison. Secondary effects are worth watching: if the attacker tries to unload the 1,783 wstETH quickly on decentralized exchanges, it could add short-term selling pressure and strain liquidity in affected pools on Base and elsewhere. Because the exploit was confined to a single vault and left Aave and Base infrastructure intact, broader systemic risk to the layer-2 ecosystem appears limited.

Why it matters

The incident is a clear data point in a wider shift in DeFi attacks. As smart contract arithmetic has matured and been repeatedly stress-tested, attackers are moving to the administrative and human layers around the code: multisig setups, whitelist management and upgrade paths. Several practical lessons follow from dev.to's account. Builders should audit permission lists and multisig configurations with the same rigor as core financial logic. Critical admin actions such as whitelist changes should sit behind time-locks, creating a window in which automated monitoring can catch and halt anomalous moves before funds shift. And depositors should not assume that an application deployed on a well-regarded network is safe, since application-level vaults can fail catastrophically even on robust chains. Analysts at ChainSentinel, cited in the report, add that tracing the post-exploit routing of stolen funds remains essential work on cases like this. Until protocol operators treat access control as a first-class security surface, permission-based exploits of this kind are likely to recur.