Techub News Exclusive: SlowMist on Kelp DAO × LayerZero Systemic Risk
This article is the full English translation of the Techub News interview.
Original link: https://techub.info/html_pages/5ccd5646-052a-4530-b6ed-4229bb9b9330.html
Interviewer: Techub News
Interviewee: SlowMist Security Team
I. Opening Question
Techub News Interview Question 1: Please define this Kelp DAO rsETH × LayerZero incident in one sentence. Is it a single-point incident, or a landmark event indicating systemic risk in DeFi in 2026?
This is one of the most severe DeFi security incidents so far in 2026, and also a concentrated outbreak of systemic risk. It is not just a case of a single contract being hacked, but a cascading failure across a three-layer architecture involving LRT (Liquid Restaking Tokens), cross-chain bridges, and lending protocols — where a single DVN configuration compromise ultimately led to losses spreading from Kelp to Aave, and then to multiple protocols holding rsETH.
Techub News follow-up question: If you had to assign one label only？
It should be “a structural problem of the entire DeFi Lego system.” The cross-chain bridge issue was merely the trigger, but rsETH being unconditionally accepted as collateral by protocols like Aave, and the lack of lending risk controls against “forged minting sources,” reflect a simultaneous failure of multiple layers of trust assumptions.
II. Background Context
Techub News Interview Question 2: Was this a code vulnerability, or a deeper trust configuration issue?
The fundamental issue here was not a coding error. The LayerZero protocol itself has no vulnerability, and the rsETH contract logic was not directly exploited.
What was truly broken was the trust configuration in the cross-chain verification mechanism — the rsETH OApp of Kelp adopted a 1/1 DVN configuration on LayerZero, meaning the security of the entire cross-chain path depended entirely on a single DVN node operated by LayerZero Labs itself. Once this node was deceived (rather than “hacked”), forged messages could pass through without obstruction.
This is essentially a “single-point trust” issue, not a “single-point code vulnerability” issue.
Techub News follow-up question: In the future, will security audits be insufficient if they only review contract code?
Completely insufficient. This incident shows that audit scope must extend from “smart contract code itself” to “cross-chain parameter configurations, DVN selection strategies, and the entire chain of trust dependencies.” An audit report that only reviews Solidity code cannot tell you how fragile a protocol is at the cross-chain layer.
III. Timeline Review: T-10 Hours
Techub News Interview Question 3: When did the attacker first show signs of activity? What happened around T-10 hours?
From on-chain behavior, the attacker carried out extensive preparatory work before executing the core attack — including preparing gas funds through mixers and pre-scanning the target attack path. This kind of planned preparation rhythm is a typical characteristic of a professional attack team, rather than opportunistic behavior after accidentally discovering a vulnerability.
The official LayerZero announcement pointed out that the attacker had obtained the RPC node list relied upon by its DVN in advance, and successfully infiltrated nodes across two independent clusters, replacing the binary files running op-geth. All of these preparations were quietly completed before the attack was executed.
Techub News follow-up question: Can this type of advance preparation be seen as having APT characteristics?
LayerZero attributed the attack in its incident statement to the Lazarus Group (TraderTraitor branch), a North Korea–sponsored APT organization. The preparation steps — preparing gas sources in advance, using mixers to avoid on-chain attribution, carefully designing the attack path so that it only returns forged data to target DVN IPs while returning normal data to others, and self-destructing malicious binaries after execution — are all typical operational patterns of highly professional APT groups, far beyond the capabilities of ordinary hackers.
IV. Timeline Review: T-0 Attack Occurrence
Techub News Interview Question 4: Please break down the most critical transaction of the attack: what exactly did the hacker do step by step, and why could the forged message pass verification?
The attack path can be roughly broken down into the following steps:
1. Infiltrating RPC infrastructure:
The attacker replaced the binary files of RPC nodes relied upon by LayerZero Labs’ DVN on Unichain, enabling those nodes to return forged on-chain state data to the DVN.
2. DDoS on normal RPC nodes:
A DDoS attack was launched against legitimate RPC nodes that were not compromised, forcing the DVN to fail over to the poisoned nodes.
3. DVN confirms forged transaction:
Based on the falsified data returned by the poisoned RPC, the DVN “confirmed” an rsETH minting/sending transaction that never actually occurred on-chain.
4. Endpoint execution release:
After DVN authentication, the LayerZero Endpoint triggered the rsETH OFTAdapter to release or mint rsETH on the target chain.
5. Cash-out phase:
The attacker used part of the obtained rsETH as collateral in lending protocols such as Aave to borrow blue-chip assets, completing monetization.
Techub News follow-up question 1: Was the most critical issue a flaw in the LayerZero framework, or a configuration issue by Kelp?
According to LayerZero’s official statement, the protocol itself operated exactly as designed. The issue lies in Kelp’s choice of a 1/1 DVN configuration —
LayerZero has explicitly marked this as a “Don’t” item in its integration documentation, and had proactively communicated best-practice recommendations to Kelp prior to the incident. From a responsibility perspective, this is a risk introduced by the integrator’s configuration decision, not a protocol-level vulnerability.
Techub News follow-up question 2: If a multi-DVN threshold model had been used, could this attack have been prevented?
As long as a second independent DVN is introduced for verification, the attacker would need to simultaneously control or deceive two independent verification systems — this would increase the cost exponentially in both technical difficulty and resources. This is also why LayerZero later stated that its DVN would refuse to sign for any application still using a 1/1 configuration.
V. Timeline Review: T+46 Minutes
Techub News Interview Question 5: From the first successful exploit to Kelp activating the pause mechanism, it took about 46 minutes. Is this response speed considered fast or slow?
Compared to many security incidents that take hours to respond, 46 minutes is not considered slow in the industry. However, for on-chain attacks, this time window is still enough to complete large-scale asset transfers, collateralization, and borrowing. The problem with DeFi is that all operations are executed within block intervals, and human intervention is inherently unable to keep up with automated attack scripts.
Techub News follow-up question: In the future, is the real solution an on-chain automated circuit breaker mechanism?
Yes. Post-incident human response can only reduce losses. What can actually stop attacks within the first few minutes is on-chain automated defense mechanisms — such as abnormal minting alerts, rate limiting on large cross-chain transfers, and automatic suspension triggered by oracle deviations. This incident should become an important milestone for the industry to push toward “on-chain automated risk control.”
VI. Timeline Review: Failed Second Wave Attempts
Techub News Interview Question 6: The hacker made several further attempts afterward, but they were unsuccessful. What does this indicate?
This indicates that the attacker’s objective was not limited to $290 million, but rather to drain as much of the cross-chain available supply of rsETH as possible. The reverted transactions suggest that Kelp’s pause mechanism took effect at the last critical moment, preventing what could have been an even larger loss.
Techub News follow-up question: If the project had responded 10–20 minutes later, would the scale of losses have increased significantly?
Most likely yes. The attacker still had an operational window before the DVN was fixed, and the exact timing of the pause mechanism directly determined the upper bound of the loss. The $290 million loss is already massive, but based on the attacker’s behavioral pattern, the number would have been higher if the attack had not been interrupted.
VII. Timeline Review: Aave Dragged Into the Fallout
Techub News Interview Question 7: After stealing the assets, why did the hacker also “incidentally trap Aave”? How did this “trap” happen?
Lending protocols are unable to distinguish on-chain between “rsETH minted through a legitimate path” and “rsETH minted via forged cross-chain messages.” For Aave, what it sees is simply a standard ERC-20 token and associated on-chain price data. The attacker deposited the anomalously obtained rsETH into Aave as collateral, borrowed highly liquid assets such as ETH, and exited, leaving behind bad debt that could not be covered by the borrowed position.
Techub News follow-up question 1: Does this expose a risk management issue in Aave, or an over-reliance in DeFi on the “authenticity of external assets”?
Both, but more fundamentally the latter. Risk parameters in lending protocols are typically based on an asset’s historical volatility and market depth, and cannot detect whether the issuance source of the asset has been compromised. This is a cross-protocol trust propagation problem that requires an industry-level solution, rather than parameter tuning by a single protocol.
Techub News follow-up question 2: Will lending protocols need to redefine what counts as “high-quality collateral” in the future?
Yes. At least for cross-chain synthetic asset categories, there is a fundamental gap between “an asset that can be priced on-chain” and “a truly high-quality collateral asset.”
In the future, mechanisms such as cross-chain origin verification and abnormal issuance monitoring may need to be introduced as prerequisite conditions for lending protocols when accepting LRT-type assets.
VIII. Structural Assessment: Systemic Risks in the DeFi “Lego” Stack
Techub News Interview Question 8: Is this the first time the full risk of the “LRT + cross-chain bridge + lending protocol” Lego structure has been completely exposed?
Yes. This is the most direct demonstration so far of composability risk in DeFi. In the past, we were discussing “a bug in a single protocol”; what was exposed this time is that when multiple protocols are connected through asset dependency relationships, the failure of any single component can propagate upstream and downstream along the value flow path, forming a cascading collapse.
Techub News follow-up question: Can we say that “DeFi appears decentralized on the surface, but is actually dependent on a few highly centralized verification points at the core”?
This assessment is quite accurate. The core issue in this incident is precisely that Kelp anchored the security of the entire cross-chain path on a single DVN operated by LayerZero Labs itself, while the LayerZero Labs DVN itself depended on a small number of RPC nodes — this forms an extremely short trust chain. In certain critical layers, “decentralized protocols” still rely on highly centralized trust assumptions, and these assumptions are often buried in documentation rather than visible at the user interface level.
IX. Technical Deep Dive: What Exactly Is DVN?
Techub News Interview Question 9: Please explain DVN in the simplest way possible, and why a 1/1 configuration becomes a fatal vulnerability.
You can think of a DVN as a “notary” for cross-chain messages. When a user wants to transfer assets from chain A to chain B, LayerZero does not directly trust the state of chain A. Instead, it requires a DVN (Decentralized Verification Network) to independently verify that “this transaction indeed happened on chain A,” and only then allow execution on chain B.
A 1/1 configuration means: only one notary is hired, and his word becomes the final judgment. Once this notary is deceived, bribed, or fed false information, the entire verification process becomes meaningless — there is no second independent voice saying, “Wait, what I’m seeing here is different.” This is the essence of a single point of failure.
X. Technical Deep Dive: Why Audits Are Still Not Enough
Techub News Interview Question 10:Many projects say they have been audited. Why do such large-scale incidents still happen after audits?
Traditional security audits mainly focus on checking whether the code logic executes as intended and whether known vulnerability patterns exist. However, the problem in this incident occurred outside the code — it happened at the “runtime configuration after deployment” layer: who performs verification, how many verifiers are required, and what happens if a verifier fails.
The industry needs to move from “code audits” to “system audits.” The latter should include cross-chain dependency configuration audits, governance permission audits, critical infrastructure dependency assessments, and most importantly — stress testing of “what is the worst-case scenario if a certain external component fails.”
Follow-up question 1: Should “configuration audits” become a mandatory separate category in the future?
I believe so. Especially for cross-chain protocols and projects using infrastructure such as LayerZero and Wormhole, DVN configurations, executor configurations, and similar parameters should be included in formal audit scope, and the audit report should explicitly disclose the security assumptions and worst-case scenarios of the current configuration.
XI. SlowMist Perspective: Tracking, Mitigation, and Industry Collaboration
Techub News Interview Question 11: From SlowMist’s perspective, what are the first actions taken immediately after a major attack occurs?
Typically, several parallel tracks are initiated at the same time:
On-chain tracing and attacker profiling:
Tracking fund flows, identifying the attacker’s on-chain behavioral characteristics, historical activity, and fund sources (including whether and which type of mixers were used), and building a structured attacker profile.
Exchange coordination:
Issuing asset alerts to major centralized exchanges, requesting monitoring and blacklisting/freezing of related attacker addresses to prevent fund withdrawal through KYC channels.
Risk alerting:
Pushing risk address lists to ecosystem participants such as DeFi protocols and wallets, helping to cut off the attacker’s subsequent operational pathways.
Whitehat negotiation channel:
In some cases, establishing communication channels with attackers and offering reasonable “bounty retention” terms to encourage partial fund returns.
Techub News follow-up question: For funds that have already gone through mixer preprocessing, what is the real probability of recovery?
To be frank, for attacks conducted by actors at the level of the Lazarus Group, full recovery of on-chain funds is extremely unlikely. They have mature fund-splitting and mixing workflows. At the current stage, the most realistic direction for the industry is to build faster cross-exchange asset freezing coordination mechanisms, and to promote stronger enforcement interfaces for on-chain evidence across multiple jurisdictions. This is currently the weakest link — the limitation is no longer primarily technical.
XII. User Perspective: Is DeFi Still Safe to Use?
Techub News Interview Question 12: After seeing this incident, the most direct question from regular users is: is it still safe to participate in DeFi?
DeFi is still usable, but users need to adjust both their risk awareness and participation methods. The core recommendations are:
You should control position size and avoid allocating large amounts of capital to highly complex products that stack “cross-chain + restaking + lending.” The more layers there are, the more risks accumulate rather than offset each other.
You should prioritize transparency and choose protocols that regularly disclose security reports and governance permissions, rather than focusing only on “high APY.”
You need to understand what you are actually holding. rsETH is not ETH — it is a cross-chain synthetic asset whose value depends on the proper functioning of an entire trust chain.
Techub News follow-up question 1: Should ordinary users avoid protocols that are “too complex to understand”?
This is a very good self-protection principle. If you cannot explain in one sentence where your assets are, who controls them, and what happens if one component fails, then that risk should not represent a large portion of your portfolio.
Techub News follow-up question 2: Should users upgrade from asking “Is it audited?” to more specific questions?
Yes. “Is it audited?” is a 2020-era standard. Today, users should at least ask: does the audit cover cross-chain configuration? What is the DVN setup (n-of-m)? What are the governance multisig parameters (m-of-n), and who holds them? These should be standard disclosures from protocols, not hidden details users have to dig out from documentation.
XIII. AI Era: Does DeFi Still Have a Future?
Techub News Interview Question 13: In the AI era, does DeFi still have a future?
DeFi not only has a future, but it may also achieve a genuine upgrade into security infrastructure in the AI era. However, this future does not belong to protocols that still rely on “complex yield structures” to attract users, but to those that are the first to introduce the following capabilities:
AI-driven real-time on-chain risk control:
Automatically identifying abnormal minting, unusual fund flows, and cascading risk signals across protocols.
On-chain insurance and automatic compensation mechanisms:
Shifting security protection from “post-incident accountability” to “in-incident coverage.”
Intelligent risk agents:
Acting on behalf of ordinary users to continuously monitor their exposure to risk, and automatically rebalancing or exiting positions when thresholds are triggered.
Techub News follow-up question 1: Will AI make attacks faster?
Yes. AI can be used to automatically scan for on-chain configuration vulnerabilities, generate optimal attack paths, and accelerate reconnaissance for off-chain infrastructure intrusions. Both offense and defense will be accelerated by AI, which means defenders can no longer rely on “manual discovery” as the last line of defense.
Techub News follow-up question 2: Will AI force the industry to shift security from a “cost center” into a “core product capability”?
This is the most promising structural shift. In the past, security investment was a “forced compliance cost.” In the future, when users begin to treat “security transparency” and “AI risk control capability” as core criteria for selecting protocols, security will become a competitive differentiator. This incident is accelerating that transition.
XIV. Closing Statements
Techub News Interview Question 14: In one sentence each, what would you say to ordinary users, founders, and the industry as a whole?
To ordinary users: Do not treat cross-chain synthetic assets as risk-free assets — every extra percentage of yield you receive is supported by trust assumptions you cannot see.
To founders: Security budgets are not an optional add-on after product launch; they are a prerequisite for whether your product can survive in a real market.
To the industry: DeFi is not dead. What we need is not more complex yield mechanisms, but more honest risk disclosure.
About SlowMist
SlowMist is a threat intelligence firm focused on blockchain security, established in January 2018. The firm was started by a team with over ten years of network security experience to become a global force. Our goal is to make the blockchain ecosystem as secure as possible for everyone. We are now a renowned international blockchain security firm that has worked on various well-known projects such as HashKey Exchange, OSL, MEEX, BGE, BTCBOX, Bitget, BHEX.SG, OKX, Binance, HTX, Amber Group, Crypto.com, etc.
SlowMist offers a variety of services that include but are not limited to security audits, threat information, defense deployment, security consultants, and other security-related services. We also offer AML (Anti-money laundering) software, MistEye (Security Monitoring), SlowMist Hacked (Crypto hack archives), FireWall.x (Smart contract firewall) and other SaaS products. We have partnerships with domestic and international firms such as Akamai, BitDefender, RC², TianJi Partners, IPIP, etc. Our extensive work in cryptocurrency crime investigations has been cited by international organizations and government bodies, including the United Nations Security Council and the United Nations Office on Drugs and Crime.
By delivering a comprehensive security solution customized to individual projects, we can identify risks and prevent them from occurring. Our team was able to find and publish several high-risk blockchain security flaws. By doing so, we could spread awareness and raise the security standards in the blockchain ecosystem.
