Generated by Codex with GPT 5.6 Sol High

Techmeme surfaced TRM Labs’ October 29 investigation, “North Korea Steals USD 6.36 Billion From Binance in the Largest Hack in History.” TRM says an attacker drained approximately \$6.36 billion from Binance-controlled wallets across ten blockchains, mostly during a 20-minute burst. It attributes the operation with high confidence to TraderTraitor, a North Korean state-sponsored hacking unit.

The most important finding is not simply the record loss. The attacker appears to have obtained valid authorizations from several parts of Binance’s signing infrastructure, then used a recent Ethereum feature to turn a familiar exchange wallet into an automated sweeper. Every blockchain accepted the transactions as properly signed even though Binance had not intended them. Cryptographic validity and institutional authorization had come apart.

A cross-chain theft compressed into minutes

TRM traces the preparation to at least 36 hours before the main theft, when an operator wallet received three small transfers from Tornado Cash. The visible attack began on the XRP Ledger at 20:05 UTC on October 29. Within four minutes, five Binance wallets sent 233.46 million XRP to one attacker address and were left with only their required minimum balances.

The operation then spread across Ethereum, Bitcoin, Tron, Solana, Arbitrum, Avalanche, Optimism, BNB Chain and Kaia. TRM values the transfers at the price of each asset when it moved. Ethereum accounted for about \$3.79 billion, including 200,630 ETH and 185 other tokens. Bitcoin outflows totaled 18,270 BTC. The breadth matters because the affected wallets used different chains, key types and signing schemes; this was not a single smart-contract exploit replicated against identical systems.

Binance’s first formal update said a compromised DevOps engineer’s laptop may have let the attacker enter a service container and submit unauthorized requests to wallet-signing services. Binance said it had found no evidence that private keys themselves leaked. A later company update said investigators had discovered additional breach points beyond the laptop and described the operation as a complex attack, so the initial explanation is not a complete postmortem.

EIP-7702 made one signature programmable

The clearest technical sequence occurred on Ethereum. EIP-7702, introduced in the Pectra upgrade in May 2025, lets an ordinary externally owned account delegate its behavior to smart-contract code by signing an authorization. The feature supports useful account capabilities such as batched transactions and sponsored fees. It also means that a valid authorization can change what an established wallet is able to do without transferring its private key.

At 20:20 UTC, the attacker submitted one transaction containing 50 authorizations. Each had been signed by Binance’s hot wallet and pointed to a sweeper contract controlled by the attacker. Three minutes and 24 seconds later, the operator told the hot wallet to send its entire balance to an attacker address. That transaction moved 167,464 ETH plus large holdings of USDT, USDC and other tokens.

The delegation remained active. As deposits continued reaching the compromised hot wallet, the attacker swept it repeatedly over the next several hours and also used it to execute other thefts. The attack therefore converted one pre-signed delegation into an ongoing control channel. Monitoring only for an exposed key or one anomalous withdrawal would have missed the more durable change in the wallet’s behavior.

TRM argues that the combined evidence points upstream to Binance’s signing infrastructure. Phishing might explain a single Ethereum authorization, but it does not readily explain why distinct wallets on multiple chains all produced valid attacker-directed signatures in roughly half an hour. Binance’s laptop-and-service-container account is consistent with that general direction, though neither Binance nor TRM has yet published the evidence needed to reconstruct exactly how the requests passed internal controls.

Freezing assets limited the loss but did not undo it

The stolen portfolio contained an unusually large amount of centrally issued stablecoins. That gave issuers a recovery mechanism that Bitcoin and Ether do not have. TRM reports that Tether, Circle, Ripple and World Liberty Financial froze about \$2.67 billion, while Arbitrum’s Security Council used its emergency authority to move 8,426 ETH and 74.6 million ARB—worth roughly another \$30 million—out of attacker-controlled addresses.

The attacker raced against those controls. About 325 million USDT was sold through professional market makers and solvers for assets that issuers could not freeze, at substantial cost. A second wallet supplied 60 million USDT to Aave and borrowed about 19,760 ETH before the affected markets were frozen. TRM estimated that the attacker still controlled approximately \$2.4 billion in assets beyond an issuer’s direct reach, including concentrated ETH holdings and BTC moving through a chain of fresh addresses.

Binance’s estimate was approximately \$6.3 billion, slightly below TRM’s transaction-time valuation. It said roughly \$3 billion had been frozen on-chain, temporarily suspended withdrawals, revoked the affected authorization and promised to cover all losses with its SAFU reserve and company funds. “Frozen” does not necessarily mean returned: recovery can still depend on issuer action, network governance, law enforcement and litigation.

The control failure happened before the blockchain

Public blockchains did what they were designed to do: they executed transactions bearing valid signatures and contract authorizations. The failure appears to have been in the system that decided which requests deserved those signatures. That distinction changes the defensive priority from protecting keys alone to protecting every service, policy engine and human workflow allowed to ask a key to sign.

TRM recommends rejecting EIP-7702 authorizations unless they are explicitly expected, treating batches of authorizations across many transaction counts as a compromise signal, and alerting when code suddenly appears at a treasury address that should behave like a plain account. Exchanges also need controls that correlate signing requests across chains. Fifty individually valid requests can look routine to isolated wallet systems while their timing and shared destination reveal a coordinated drain.

Market makers, solvers, bridges and lending protocols form a second defense layer. In this case, some counterparties exchanged hundreds of millions of dollars for a wallet funded by the theft only minutes earlier. Screening at transaction speed cannot guarantee recovery, but it can reduce the attacker’s ability to convert freezable assets into censorship-resistant ones before labels propagate.

What remains uncertain

TRM’s transaction reconstruction is unusually detailed, but its attribution is still an intelligence assessment. The company says robust on-chain and off-chain evidence supports TraderTraitor and notes that the FBI attributed the 2025 Bybit theft to the same unit. It does not publish the off-chain evidence behind this new judgment, and no government attribution is cited for the Binance incident. TRM also sells blockchain-intelligence services, an interest readers should keep in view even though much of its transfer analysis can be checked against public ledgers.

The root cause is likewise provisional. Binance’s initial explanation focused on one compromised laptop and one service container; its follow-up acknowledged additional breach points, and a full technical postmortem is still pending. Valuations and frozen totals vary with asset prices, the time of measurement and later interventions. The event establishes that a coordinated attacker obtained extraordinary signing authority, not yet exactly how every control failed or who operated the infrastructure behind it.

Even with those limits, the incident provides a clear engineering lesson. A hardware-protected private key is not a sufficient boundary when surrounding software can request powerful, programmable signatures. Security teams need to model what a valid signature can delegate, aggregate intent across chains, and make abnormal authority changes visible before the first large transfer—not after the ledger has faithfully finalized it.