The code does not lie; only the auditors do.
On May 22, a series of transactions ripped through the Wildberries protocol—a supposedly “civilian” cross-chain logistics aggregator that had quietly moved over $2.1B in volume since its launch in Q4 2025. The exploit wasn't a flash loan attack. It wasn't a reentrancy trick. It was something far more surgical: a targeted destruction of the protocol's core relay nodes, executed by a wallet cluster that had been dormant for 400 days.
Volume is vanity; on-chain flow is sanity. I traced the flow, and I traced the lies. What I found was a blueprint for how a DeFi protocol's “non-military” infrastructure can become the perfect kill box for a coordinated on-chain raid.
CONTEXT: The Protocol That Wasn't Supposed to Be a Target
Wildberries markets itself as a neutral settlement layer for decentralized physical infrastructure networks (DePIN). Its value proposition is simple: relay data between disparate blockchains without relying on a centralized oracle. It processes logistics data—shipping container locations, temperature readings, delivery timestamps—and anchors them on-chain via a network of “relayers.” These relayers are analogous to Wildberries' physical warehouses: they store, verify, and forward data. The token, $WILD, is used to stake into relay slots and earn fees.
On paper, it's a textbook DePIN project. The team has a BS in software engineering from a top university. The whitepaper is 47 pages, well-written, with formal verification claims from a mid-tier auditing firm. The CEO gave a TEDx talk titled “Bridging Atoms and Bits.” The narrative was flawless.
But I don't read whitepapers. I read bytecode.
CORE: Systematic Teardown of the Relayer Destruction Attack
The Hook – A Dormant Wallet Cluster
On May 22, I was monitoring the Wildberries relayer registry contract (0x7F3c…AB12) for unusual activity. My Python script flagged a single transaction that modified the relayer eligibility for the Avalanche sub-net. The caller was a fresh address funded by a Tornado Cash withdrawal from 2024. That should have been the first red flag—but the protocol's own dashboard showed no alert.
I do not guess; I verify.
Step 1: Reconnaissance of the Registry Logic
I decompiled the registry contract. Wildberries had implemented a custom “emergency admin” function called updateRelayerSet that allowed a multisig of 5 addresses to modify relayer permissions without a timelock. The function had a visibility modifier of external, meaning any contract could call it if the caller matched the multisig threshold.
But here's the flaw: the multisig check used a msg.sender equality for each signer, not a cryptographic signature verification. The protocol assumed the multisig wallet would always be a Gnosis Safe, which enforces internal signature validation. However, the contract itself did not verify that the msg.sender of each signer was the same address that had previously approved the transaction. The check was:
require(signers[0] == msg.sender || signers[1] == msg.sender || …);This is a classic “address equality vuln”. The emergency admin never envisioned that the multisig wallet itself could be compromised. The attackers had to simply acquire the private keys of three of the five signers. How? I traced the origins of the signer addresses.
Step 2: The Key Theft
Signer #2 (0x8B3d…F90E) had been deployed by the same deployer as the Wildberries team's main contract. That address had also minted 500,000 $WILD tokens to a personal wallet that was later used to create liquidity on a decentralized exchange. The wallet's seed phrase had been accidentally committed to a GitHub repo in Q2 2025—a repo that the team said was “internal only” but was actually public for 8 days. I found the commit hash via a Google dork.
Signer #3 (0xA9c1…E4D2) was a known address that had interacted with a phishing site in early 2026. The phishing site was masquerading as a Wildberries governance portal. I found evidence of a token approval transaction that drained the signer's $WILD stake, but the signer had never revoked permissions. The attacker didn't drain the stake; they copied the private key.
Signer #4 (0xF2e8…C7B1) was the most damning: it was a vanity address generated using a tool that had a backdoor. The vanity address was created by an outsourced developer who still had access. That developer sold the key for 50 ETH in March 2026.
Silence is the loudest admission of guilt. The team never disclosed these compromises.
Step 3: Execution – The Relayer Purge
With three keys, the attackers called updateRelayerSet and replaced all honest relayers with addresses from their cluster. They then triggered a pauseAllRelays function that was meant for emergency maintenance. But the function had a secondary effect: it allowed the admin to withdraw all pending fees from the relayers. The attackers drained approximately 12,000 ETH (worth ~$38M at the time) from the pending fee pool.
The protocol's frontend showed “Scheduled Maintenance.” No one noticed for 47 minutes.
I did. I was watching the mempool and saw the fee drain. I immediately published a warning on my Telegram channel. By the time the team reacted, 80% of the drained funds had been swapped to ETH and bridged to a private chain via a cross-chain aggregator.
Step 4: The Aftermath
The $WILD token dropped 73% in 6 hours. The Wildberries team issued a statement blaming “a sophisticated state-backed attack” and promised to reimburse users through a new token. But I checked the team's own multisig: they had moved $WILD tokens to a new contract 3 days before the attack, right after the audit report was published. The code does not lie; only the auditors do.
The “audit” from CertiK (cough) covered only the core swap contract, not the registry or the emergency admin. The team relied on a “partial audit” narrative to deceive investors.
CONTRARIAN: What the Bulls Got Right
Let's be coldly objective. The Wildberries concept itself is not worthless. The idea of a neutral logistics relay layer has genuine utility for supply chain finance. The protocol's on-chain volume growth was organic—I verified that 72% of transactions came from unique addresses with non-zero gas usage patterns, not wash trading. The team's initial technical execution was sound: the cross-chain relay latency was under 2 seconds, which is genuinely impressive.
But here's the hard truth: the bulls ignored the human factor. They saw the TEDx talk, the BS in Software Engineering, the formal verification claims, and assumed the code was airtight. They never asked: who holds the keys to the emergency admin? They never ran a simple Google dork for private key leaks. They never checked if the vanity address generator had a backdoor.
In military terms, the bulls assessed the front-line defenses (the smart contracts) but ignored the logistics tail (the key management, the developer credentials, the outsourced code). The attack was not a technical failure of the protocol's core logic—it was a failure of operational security (OpSec).
The bulls argued that Wildberries was “too big to fail” and that the team would recover. But I have seen this pattern before: when a protocol's key management is compromised, the trust is irreversibly broken. No amount of token redemptions can restore the confidence that the same vulnerability won't resurface. The only recovery path is a full re-deployment with a new key structure, which is effectively a restart.
TAKEWAY: Accountability Requires More Than Code
Every transaction leaves a scar on the ledger. The Wildberries scar is now part of the permanent on-chain record. The attackers are still anonymous—that's the nature of our industry. But the real failure was not the hack itself. It was the illusion that a protocol can be “neutral” and “trustless” while its operational layer runs on a handful of private keys held by fallible humans.
The next time a DePIN project boasts about its “decentralized logistics,” demand to see the multisig signer list. Demand a public audit of the OpSec procedures. Demand that the team prove they didn't accidentally commit a seed phrase to a public repo.
Promises are encrypted; data is decrypted. I traced the flow. You should too.