$6 million out of a Base vault: what you can do now with your DeFi deposits

An investment vault on the Ethereum layer Base was drained of around 1,783 wstETH on Sunday, October 4, 2026, worth about $6 million at the October 5 price. What mattered was not a hole in a protocol but an entry on a permission list: the attacker got a freshly deployed contract onto the vault's whitelist, borrowed interest-bearing deposits through it and redeemed them at Aave V3 for Ether derivatives.

For an investor in Germany this is not a distant protocol accident. Anyone with funds sitting in a vault, a yield product or a lending strategy depends on exactly the permission structure that failed here. This piece sets out the sequence, names the places where control was missing, and sorts out what a German investor can look at in their own set-up.

What happened in the Base vault on October 4: 1,783 wstETH on a Sunday morning

The incident began on the morning of October 4. At 08:52 UTC the Safe wallet of the vault owner removed a contract belonging to the attacker from the whitelist. One minute later, at 08:53 UTC, the same contract was admitted again. Both transactions carried valid signatures from existing signers.

At 09:20 UTC the first damage estimate from the security firms watching the chain stood at $2.02 million. Over the course of the morning the sum grew to around $6 million, because the outflow continued. In the end 1,783 wstETH sat in the attacker's hands, transferred to an address they controlled.

The arithmetic can be followed: wstETH was quoted at around $3,386 on October 5, with Ether itself at about $2,720. At 1,783 units that comes to a good $6 million. The figures given in reporting range from $6 million to $6.1 million, depending on the price moment used. Nothing is being smoothed here: the amount hangs on the price, and the price moves.

Several security firms watched and analysed the sequence at the same time, among them Blockaid, PeckShield, CertiK and ExVul. That simultaneity is the reason the event is documented minute by minute even though the vault itself has issued no statement.

The permission list as the way in: how a new contract reached the funds

A whitelist is a list of addresses or contracts that a protocol admits for a particular action. Anything not on the list is rejected. The method counts as solid as long as one condition holds: whoever may change the list has to stay the same party.

That is exactly where the attack went in. The attacker had to outwit no calculation rule and find no weakness in Aave's code. All they needed was a place on the permission list for borrowing. With that place their contract behaved, as far as the system was concerned, like any other admitted contract.

The difference between a protocol flaw and a permissions flaw

A protocol flaw exists when the calculation logic of a contract gets its arithmetic wrong, for instance in a price calculation or an interest formula. A permissions flaw exists when the logic works correctly but the wrong party is allowed to operate it. The Base vault is a case of the second kind, and that kind is the more uncomfortable one for depositors: an audit that confirms the calculation logic says nothing about how the permission list is managed.

There is a further detail that shows up in several analyses: before October 4 the vault's Safe wallet had triggered no transaction for about 25 days. A container holding millions that stays inactive for weeks has nobody watching when one suddenly does appear.

08:52 and 08:53 UTC: the sequence inside the Safe multisig, minute by minute

The short run of the two whitelist transactions is the strangest part of the event. A removal followed by a re-admission sixty seconds later does not look like an operating error. The sequence reads more like a process in which somebody deliberately set the order of the approvals.

The analysing firms consider two explanations possible, and both are phrased as possibilities rather than findings: either keys of existing signers were in somebody else's hands, or the approval path itself was manipulated, so that signers confirmed something other than what they believed they were confirming. Which of the two applies is open. No statement from the operator is available.

Only the technical side can be stated for certain: the signatures were cryptographically valid. For the blockchain the change to the permission list was therefore a compliant action. No rule was broken, and nobody could have rejected it on the basis of the signatures.

Valid ECDSA signatures: what a three-of-seven multisig does and does not do

According to the on-chain analyses the vault hung on a Safe multisig with a threshold of three out of seven. A multisig is a wallet in which an action is carried out only once a set minimum number of signers agree. Three of seven means any three of the seven stored keys suffice for every approval.

This design protects against the loss of individual keys. If one key fails, the wallet stays usable. Against another risk it offers little: whoever holds three of the seven keys together has full control. With seven signers whose identity is not publicly known, it is not even possible to judge from outside how independent those seven are from one another.

Why the threshold on its own says nothing

A multisig is as strong as the spread of its keys. Seven keys in seven hands in seven places are a different proposition from seven keys sitting on two devices in one office. From outside the two look identical: the chain shows only the threshold, not the distribution. If you set up a multisig yourself, the distribution is the part that does the work; the procedure is in our guide to setting up a multisig wallet.

The second point concerns the delay. Many protocols place a lock-up period, a timelock, between a change of permissions and the moment it takes effect. A change to the permission list then becomes valid only 24 or 48 hours later. In the case of the Base vault there was no such period between approval and outflow. Had there been one, the other side would have had hours rather than seconds to react.

For your own holdings that means this: custody outside such containers is the part you alone decide on. Which devices come into question and what they cost is set out in our hardware wallet comparison.

Aave V3 and the Base chain: why this was a failure of access control

Neither Aave's core contracts nor the Base network itself were compromised. Both worked as intended throughout the event. The attacker used Aave V3 as a payout desk: they borrowed interest-bearing deposits and swapped them there into wstETH, which any admitted contract is allowed to do.

That distinction matters for the assessment and is often blurred in fast reporting. An attack on Aave would have affected every user of the protocol. What happened here affected a single vault and its depositors. That does not make the damage smaller for those caught in it, but it changes the question an investor has to ask. What counts then is who controls the container you paid into, and only secondly how safe the large protocol behind it is held to be.

What the analysts say is still sitting in the vault

According to on-chain analyses an eight-figure dollar sum remained in the vault after the outflow, under the same permission structure as the part that left. Figures quoted are around $31.7 million. Since the operator has not commented and no recovery measure has been announced, that figure remains an outside estimate. Part of the proceeds was bridged to Ethereum over Lido infrastructure according to observations from the developer community; that too is an observation, not a confirmed attribution.

wstETH, aBaswstETH and Lido: the token wrappers in this case explained

Three names turn up in this incident, and they build on one another. Keeping them apart makes it clear why the outflow needed two steps.

stETH is the claim on Ether staked with Lido. wstETH is the wrapped version of it, fixed in unit count; the yield is reflected in the price instead of in the quantity. aBaswstETH, finally, is the claim on wstETH deposited with Aave on Base and earning interest there.

The attacker borrowed the third form and redeemed it into the second. Each of these wrappers is tradable, and each carries its own risk: a wrapper can lose its peg to the underlying, and a wrapper sitting on another chain also depends on the bridge that took it there. Anyone generating yield through such stacks is better off counting the layers before looking at the percentage.

A vault without an operator: what an unnamed issuer means for depositors

The most striking point about this case is not in the technology. Up to the publication of the first analyses no protocol had claimed the vault and no party had announced a recovery. So there are funds in the millions, a multisig with seven unknown signers and no office an injured party can turn to.

For German investors that is the decisive legal difference from everything that has fallen under the EU regulation MiCA since January 1, 2026. An authorised provider of crypto asset services has a registered office, a supervisor and an address for complaints. A vault without a named issuer has none of that. Which providers hold an authorisation in Germany can be looked up in BaFin's company database. Many readers know the permission list as a component from the user side: a withdrawal whitelist in your own exchange account works on the same idea, except that there you decide yourself who is on it.

Yield against counterparty risk: the difference between a DeFi vault and a licensed provider

A vault usually promises a higher return than a staking offer at a supervised provider. That premium is not a gift. It is the payment for risks the depositor takes on: the risk of the permission structure, the risk of the bridge, the risk of an operator who in case of doubt cannot be reached.

At a supervised provider the risk sits elsewhere. There you carry a counterparty risk against a named company, but with authorisation, reporting duties and a route for complaints. Which version fits better depends on how much you can and want to check. Supervised routes to staking returns usually pay less, but a named company stands behind them.

The case fits a pattern of the year. In September 2026 more than $766 million disappeared from exchanges, bridges and DeFi protocols according to industry counts, the highest monthly figure of the year. Among the larger incidents of the year the cause lay predominantly in keys and permissions, and only after that in the calculation logic of the contracts. Two days before the Base vault, on October 2, two Safe multisigs with an activated loop module around Aave v3 were hit, out of which 114.09 ETH flowed; there too the core pools were untouched, and there too the fault lay in a helper contract with permissions that were too broad.

A loss from a vault outflow: the tax position in Germany

For tax purposes the theft of crypto assets in Germany is not a clearly regulated case. Under the law as it stands a sale triggers a private disposal under Section 23 of the Income Tax Act. An outflow through third-party access is not a sale, so the event a loss could attach to is missing.

In practice that means anyone wanting to claim such a loss needs records. Those include the transaction hashes of the deposit, proof of the moment of the outflow, the address of the vault and documentation of your own acquisition costs. Whether the tax office recognises the loss is decided case by case; there is no general assurance. Anyone affected for larger sums should settle it with a tax adviser, not with an article.

What you can document while the chain still shows it

Blockchain data does not disappear, but analytics services change their presentation and providers switch interfaces off. Anyone affected should save transaction records and account statements to a file promptly instead of relying on a service's interface. It costs half an hour and cannot be made up later.

Base vault hack: the key points for your decision

1. Look up who holds the rights to your container. For every product holding funds the question is the same: who may change the permission list, how many signers does it take, and is there a lock-up period between a change and the moment it takes effect? If you find no answer to that, the answer is the absence. For the part you want to hold yourself, our software wallet comparison helps with the classification.
2. Separate yield balances from holdings. What stays put for the long run does not belong in a container with somebody else's permission structure. Anyone generating returns through lending will find the supervised routes in the lending comparison; they usually pay less, but a named company stands behind them.
3. Keep the purchase trail traceable. You need the time and the cost of acquisition for any tax classification, including a loss event. Providers authorised in Germany that issue clean records are listed in our overview of crypto exchanges.

The Base vault remains for now a case without a sender. What it shows is useful regardless: the question of how safe a protocol is differs from the question of who controls your own funds. No audit answers the second question; its answer is in the list of signers.

(As of October 5, 2026. This article is not investment advice. Prices and fee structures change; check the terms with the provider before you buy.)

Sources on the sequence: The Crypto Times on the whitelist access and Tokenpost on how the damage figure developed.