DeFi won, but at what cost?

Context

On April 18th, the North Korean Lazarus hacker group compromised KelpDAO’s rsETH by exploiting the LayerZero bridge through a combination of RPC poisoning and targeted DDoS, and redeeming 116,500 unbacked rsETH tokens on Ethereum. KelpDAO had deployed a 1-of-1 DVN configuration, making a single compromised verifier sufficient to forge a cross-chain message.

Rather than selling the stolen funds, the attackers deposited them as collateral on Aave, Compound, and Euler, borrowing approximately $292 Million in ETH across several protocols with the majority of the funds being borrowed on Aave across multiple networks.

→ The full technical breakdown of the exploit mechanics, attribution analysis, and the role of each party, is available here:

Since the hack, the total DeFi TVL declined by 17% with Aave being hit the hardest with over $12 billion outflows, around 47%.

---

Community in the dark, lenders stuck

Once the hack consequences were clear, the resolution of the situation was still not over and the spillover effect only continued. With DeFi users pulling their funds out of Aave and the emode enabling rsETH collateralisation for ETH borrowing, lenders found themselves stuck on Aave with little to no liquidity left to withdraw from.

This is a feature of the pooled lending system: lenders deposit their funds and borrowers take these funds out by providing collateral. However, until borrowers repay their loan, lenders might find themselves out of withdrawable liquidity.

In order to incentivize borrowers to repay their loans, or attract new depositors on the protocol, Aave uses the interest rates on its lending pools. If the pool utilization is high, these rates spike, making loops and yield farming strategies unprofitable, while increasing the payout to lenders.

Here is a more technical explanation for those willing to go more in-depth. If you want the practical explanation, you can skip this part. The IRM (Interest Rate Model) is divided into two distinct zones by a pivot point known as the "kink" (denoted as U*).

• Before the kink (0% → U) : The curve is "gentle." This is the optimal operating zone where rates rise slowly to attract depositors without overly penalizing borrowers. The slope of this section is called Slope 1.*
• After the kink (U → 100%) : The curve steepens sharply. This is the "stress zone," signaling to borrowers that capital is becoming expensive and to depositors that they can earn a premium for providing liquidity. This slope is called Slope 2.*

Currently, stablecoin pools are stuck in the Slope 2 region, with U* set at 92% for USDC. Beyond this threshold, the deposit APY spikes to incentivize liquidity inflow, while borrowing rates jump to discourage further debt.

This scenario might work in a “normal” market situation where DeFi users and institutions are looking to get additional yield, but not under stress conditions where users are not certain to keep their initial capital or become illiquid.

In response to the liquidity freeze, Circle's Chief Economist Gordon Liao put forward a proposal to aggressively hike Aave's interest rate parameters - raising Slope 2 to 50% and lowering the kink to 85% - which would push maximum borrow rates to 53.5% and deposit rates to 48.2%, in theory pressuring borrowers to repay while attracting fresh liquidity from the outside.

The community, however, was unconvinced. At 48% APY, the yield is eye-catching, but when confidence has collapsed, no rational actor ties up capital for a return that comes with no exit guarantee. Worse, the abrupt shock could detonate over $70M in vulnerable looping positions before a single new depositor steps in.

The proposal also cast an uncomfortable light on Circle's posture relative to Tether, which had responded to the recent Drift protocol crisis by extending a direct $100M credit line - a very different kind of commitment, and one the community has not forgotten.

---

The DeFi United relief vehicle

In response to the hack, Aave and KelpDAO started a DeFi United initiative, a relief vehicle to replenish the rsETH reserves and make all the holders whole. With a coordinated outreach to partners and affected parties, they managed to plug the majority of the hole at the moment of publication.

This part will be updated as new donations come in, and finalized once the relief vehicle has fully replenished rsETH holders and that the situation is back to normal.

The disclosed donations

Protocol donations

• Lido (2,500 ETH)

Lido was the first protocol to announce its participation in the relief vehicle with a donation of up to 2,500 stETH.

The key exposure Lido is trying to protect is its EarnETH vault that holds strategies linked to rsETH. If the relief vehicle is not fully funded, EarnETH vault depositors could face up to 9,000 ETH in losses. So Lido is contributing to avoid a larger mess for its own users.

One important condition is worth highlighting here: Lido's funds only deploy if the full deficit is covered. The 2,500 stETH is explicitly contingent on other contributors closing the entire 120,000 ETH gap. Lido refuses to be a partial backstop that still leaves EarnETH users exposed, meaning it's all-or-nothing participation.

On top of this 2,500 stETH contribution, Lido is also contributing up to $3 million of its own shares in the vault to be burned should the relief vehicle fall short.

So far, the governance proposal is likely to get passed with over 12 million votes in favor of helping the rsETH relief vehicle.

• EtherFi (5,000 ETH)

EtherFi proposed to use its treasury to restore the collateral assets of rsETH by injecting 5,000 ETH into the rsETH relief vehicle to cover the collateral gap, protect user assets, and prevent bad debt from spreading across DeFi.

It is unclear if this announcement’s goal is to take market share from its competitor, but this represents one of the largest donations from a non-affected protocol.

• Golem Foundation & Golem Project (1,000 ETH)

Golem Foundation posted their announcement allocating 1000 ETH to the relief vehicle to support the broader DeFi ecosystem. They had no exposure to rsETH and from our understanding appear to have contributed purely out of goodwill.

• Aave DAO Treasury (25,000 ETH)

Aave Treasury committed to donating 25,000 ETH to the DeFi United effort through a governance vote proposed by Token Logic. We expect this vote to pass, and therefore have included this donation directly in the disclosed donations section of the report.

This is the clearest full accounting of the rsETH crisis we've seen so far from one of the 3 involved parties.

Individual donations

While protocols stepped up to cover the hole, a number of individual donors and unaffiliated entities stepped up with personal contributions.

• Stani Kulechov, Aave founder and CEO @ Aave Labs (5,000 ETH)
• Ernesto Boado (CEO @ BGD labs) and BGD Labs (350 ETH)
• Emilio Frangella, VP Engineering @ Aave Labs (500 ETH)

An address was also created for public donations (defiunited.eth) that can be tracked here.

The Mantle loan

The Mantle foundation decided to step in to help plug the hole. It is quite interesting because Mantle is not transferring money to the relief fund, but instead decided to lend up to 30,000 ETH to this initiative.

Here are the specifics:

• Loan amount: Up to 30,000 ETH
• Interest rate: Lido rate + 1% (Ethereum staking rate with interest on top)
• Maturity: 36 months maximum with early repayment possible with no penalty.
• Conditions for Aave: 5% of total Aave DAO revenue and a collateral of $11 million in AAVE tokens. On top of this, Aave will delegate 130,000 AAVE tokens for Mantle to participate in the protocol governance.
• Custody: multisig wallet with Mantle holding a key to this wallet.

The undisclosed donations

These are the protocols and entities that committed to donating an undisclosed amount. This section will be updated as we learn more about their commitments and they will be moved to the protocol donations once the final contribution has been announced.

• Ethena Labs

It made sense for Ethena to contribute to this campaign as USDe and sUSDe loopers on Aave are directly affected by the lack of liquid stablecoins and are currently experiencing negative returns.

?s=20

• LayerZero

LayerZero, whose infrastructure served as the attack vector, has announced that they will also be contributing to the DeFi United initiative shortly after Ethena’s announcement. Their contribution should, in theory, match or exceed the largest donations so far as their infrastructure was involved. At the moment of publication, they have not disclosed the final amount allocated to the relief vehicle.

• Ink Foundation & Tydro

Tydro is a “friendly” Aave fork deployed on Ink. Along with Ink Foundation, they have announced public support for this initiative without providing further details at the moment of publication. We could see something similar to what Mantle proposed on Ink’s side as they would heavily benefit from Aave’s success and their further integration in incentive programs and deeper liquidity.

• Frax Finance

While Frax did not disclose direct participation, we are including their statement here to acknowledge their support. We believe Frax might contribute to deeper liquidity in borrowable stablecoins, allowing users to exit their positions in a more efficient way.

• KelpDAO

It is worth mentioning that KelpDAO has not committed to anything yet. We reached out to them and received a confirmation that there will be a financial commitment on their end for this initiative. The amount is not yet disclosed.

The Arbitrum Freeze

Besides the DeFi United relief vehicle, a portion of funds were also successfully recovered on Arbitrum. On April 21, Arbitrum announced that it had frozen 30,766 ETH linked to the Kelp DAO hack, representing approximately $71 million. The funds were transferred to a secure intermediary wallet and can no longer be moved without a new governance decision.

They did not perform a rollback or alter the network’s history to recover the funds. Instead, the intervention was executed through a special transaction at the system level, known as an ArbitrumUnsignedTxType.

This type of transaction cannot be produced by a regular user, even with their private key. It is an operation reserved for ArbOS (Arbitrum’s operating system) and can only be executed by the sequencer, under the requirement of a multisig validation by the Security Council (12-of-N model), limiting its use to exceptional cases. Here, 9 out of 12 members of the Security Council successfully executed the transaction to freeze the ETH on its chain.

---

Why it worked and why it might not work again

Now let’s face the facts. So far the DeFi United relief vehicle worked because… well, it was championed by Aave. This was only possible because Aave has a huge network effect with partners that also had enough exposure to this disaster and had incentive to help plug this hole.

This situation has an eerie resemblance to what happened on Aave in 2023 with Michel Egorov, the founder of Curve Finance.

A Déjà-vu?

If you are familiar with the situation that took place in 2023, you can skip the next 3 paragraphs. For those that need to be brought up to speed, here is a recap:

On July 30, 2023, Curve was exploited due to a vulnerability in the Vyper programming language, resulting in around $70 million being stolen. This raised concerns about Egorov's collateralized loans: he had borrowed over $100M against approximately 460M CRV tokens, representing 47% of the total supply. A drop in the price of CRV could lead to a liquidation of his position and kickstart a liquidation cascade.

His largest exposure was a $63M loan from Aave, collateralized by 34% of the CRV supply. The problem was structural: in the event of liquidation, Aave would have had to sell the CRV put up as collateral to the open market, which would have had a cascading effect due to a lack of liquidity.

The resolution came through informal OTC deals: in August, Egorov sold 106 million CRV for $46 million to reduce potential liquidation risks, with buyers including crypto trading firm Wintermute, Tron founder Justin Sun, and NFT investor Jeffrey Huang, as well as Aave that bought $2 million of CRV tokens. An anonymous entity secured the largest OTC deal, purchasing 17.5 million CRV tokens.

Unfortunately Egorov did not return the favor for this hack. He responded on Twitter that his participation would be “not easy without selling [his] kidney” and that Curve didn’t have funds for it.

The resemblance lies in the whole DeFi ecosystem rallying together to solve a crisis. It even has somewhat of a poetic connotation to it. It truly feels like the ecosystem comes together to resolve a crisis affecting not only users but also the most important protocols in the industry. These moments make DeFi feel like an industry worth fighting for.

One factor worth examining is why this response was so effective.

Aave is the success factor in this recovery

All the success that the relief vehicle has had so far can be in large part attributed to Aave, with KelpDAO and LayerZero largely benefit from Aave’s touch.

This “touch” refers to the composability Aave has with other protocols that depend on Aave working properly, the communication with slogans such as “DeFi United”, “Just Use Aave”, or “Aave Will Win”, the coordinated communication once the funds were secured, as well as Aave’s importance for the crypto ecosystem.

DeFi won because Aave was the protocol affected.

The response might not have been the same if Kelp was not part of Aave and only existed on some protocols that were not this important for the ecosystem.

This also explains why Aave did not have to resort to Umbrella’s module, therefore punishing its users that trusted Aave with their judgement even if they knew they would be the backstop in case Aave didn’t succeed to plug the hole.

This explains why users did not get a haircut on their holdings. Aave has to save face to remain reliable and maintain the image of “no ghost left behind”. Aave has to succeed for DeFi to survive.

This also explains exactly why Aave had a larger involvement compared to LayerZero or Kelp in public communications. LayerZero could have ended up like another bridge hack. Kelp could have ended up like a restaking protocol that did not configure their crosschain solution properly. Aave just wasn’t allowed to fail or lose the trust of their users.

Here is the most blunt way we can put it: Aave’s solution to backstop this type of situations, the Umbrella module, should be the last resort for the protocol to maintain its users’ trust.

However, this does raise some serious questions about our industry and what we are doing here.

Some questions worth asking

The rsETH hack proved DeFi can survive a $292M exploit against its dominant lending protocol. What it also proved is how many specific conditions had to align for that survival to work. Aave had to be unusually well-capitalized. The attacker had to leave funds on a chain willing to freeze them. The adjacent ecosystem had to be scared enough of contagion to contribute.

• What does “De” in DeFi stand for?

The rsETH recovery stress-tested that question harder than any other event in DeFi's history. Arbitrum's Security Council froze an attacker's funds by executive decision. Aave did not activate its Umbrella module designed to backstop exactly this type of situation.

The "community" that mobilized treasury assets and approved relief frameworks is, in practice, a small number of large stakeholders acting rationally in their own interest. It may be the only realistic way a young financial system survives existential shocks.

But it should retire the comfortable fiction that DeFi's resilience comes from decentralization. It comes from financial depth, concentrated coordination capacity, and the willingness of large players to act. The "De" was always an aspiration; in practice, execution looks very different

If Aave fails, DeFi may fail. If the entire DeFi ecosystem relies on a single protocol, is it really “decentralized”?

The reality is that for years, we at OAK Research have argued that we should stop talking about “DeFi” and instead refer to “on-chain finance,” because aside from a few isolated parts of the ecosystem, nothing is truly decentralized anymore. If there’s one positive takeaway from this hack and the recovery process, it’s that it may help more people confront this reality.

• Will Umbrella ever work as intended?

Umbrella held $54 million when the protocol needed around $230 million. This means that even if Umbrella were to be activated, the shortfall would still be around $175 million to cover.

The deeper issue is the activation of this module: what we’ve seen is a large exodus of capital once the impact was clearly announced. So far neither the Safety Module (the predecessor of Umbrella), nor Umbrella were ever used. And there is a reason for it: once this module suffers a loss, no user would take the risk to lock their funds in it, effectively eliminating the security buffer Aave has had so far.

At the same time, as we’ve seen here, Umbrella is the last resort for Aave to use. If a user gets paid a small premium to suffer a total loss of capital, market confidence will erode. Aave will resort to any possible solution (even draining its own treasury) before using the Umbrella funds.

Umbrella will work as intended exactly until the moment it is tested at scale, which is the same paradox that makes every insurance system in the ecosystem undercapitalized at the worst possible time.

• What’s the path forward for Arbitrum?

The Security Council's intervention was decisive, effective, and deeply uncomfortable. It saved $71 million. It also demonstrated that Arbitrum One is, at its core, a system where nine signatures can override the chain state, which is a different product than the one the decentralization narrative sells. Even though Arbitrum is not a Stage 2 rollup yet, it has created a precedent that will be difficult to overcome.

This move was a legitimate and arguably great protection mechanism that has placed Arbitrum in an uncomfortable position, potentially setting a legal precedent for future incidents that might happen.

From our point of view: Arbitrum has to accelerate its path to a stage 2 rollup as soon as possible in order to eliminate this precedent and insulate itself from liability in future incidents on its chain.

• What happens for the next hack that isn’t on Aave?

The rsETH response worked because the victim was DeFi's largest protocol, with $181M in treasury assets, $140M in annual revenue, and a gravitational pull large enough to make every adjacent protocol's survival contingent on its own.

Apply the same attack to a mid-sized lending protocol, a smaller ecosystem, a chain without a Security Council, a vault curator, and the outcome looks very different

No one runs to its rescue. No ecosystem fund mobilizes. No law enforcement coordination freezes the funds in time.

Aave might have become the only protocol too big to fail with enough resources and a strong network effect that allowed it to survive this hack. Any other protocol would not be in the same position today.

Conclusion

This conclusion is a personal opinion of the writer of the article and should be taken as such.

If it was not clear enough, I strongly believe that if the same situation took place elsewhere but Aave, the outcome would not have been the same.

I would like to commend the current leadership at Aave for coordinating these efforts and showing their commitment to the protocol. 30.6K ETH came from Aave or Aave-affiliated individuals and entities.

• Aave DAO: 25,000 ETH
• Stani Kulechov: 5,000 ETH
• BGD Labs and its founder Ernesto: 350 ETH
• Emilio: 500 ETH

These people alone have contributed over 40% of the total amount needed to plug the hole. They have coordinated the efforts and reached out to people to make sure no user gets a haircut on the funds deposited on Aave (or, for that matter, elsewhere). It bears saying: the people leading this protocol have definitely proven that Aave is the leader of the ecosystem for a reason.

At the same time, the response from the two main actors that were affected by this hack, LayerZero, and KelpDAO, was quite disappointing. While the ecosystem rallied together, both of them kept pointing fingers at each other, and no concrete contribution on their end was announced at the time of the publication. Aave owned up to its mistakes. This cannot be said for these protocols. While both had earned significant trust in the crypto ecosystem, this episode revealed how much the outcome depended on Aave rather than them.

What this episode has shown is that the industry is filled with single points of failure, be it within individual protocols or across DeFi as a whole, where a single protocol has become too big to fail. Is composability really needed, and at what point is it worth taking the risk of integrating exotic assets into a protocol?

As always, the DeFi ecosystem only learns once something breaks. Here, a lot of things broke down, and there will be many changes coming to the ecosystem to make it more reliable, secure, and more professional overall. At least until the next time something happens.