Binance resumes withdrawals after $6.3bn hack: a 20-minute attack and lessons for payments engineers
After the record 29 October attack, Binance reopened withdrawals in phases from 1 November, with $3.64bn in net outflows on day one. What it teaches about hot wallets, limits and production access.
What happened
On 29 October 2026, about $6.3 billion in crypto was stolen from Binance's hot wallets, the online wallets used for everyday payouts. According to TRM Labs, outflows started at 20:05 UTC, and roughly $6.36 billion left across ten blockchains, mostly within about 20 minutes. It is the largest known crypto theft; the 2025 Bybit hack involved about $1.5 billion.
Binance's initial investigation linked the attack to a compromised DevOps engineer's laptop, which let the attacker reach an internal service container and submit unauthorized withdrawal transactions. Co-CEO Richard Teng then said the attacker "did not simply use one engineer's permissions" and that the investigation "has not attributed responsibility to any individual." Binance said it had uncovered "additional points of breach" and found no evidence of a wallet private key leak.
On attribution: TRM Labs and Chainalysis attributed the attack to TraderTraitor, a North Korea-linked group, and the FBI later reached the same conclusion. That is an assessment by investigators and analysts, not a court finding; Binance says it will publish a detailed report once its investigation is complete.
Binance froze all deposits and withdrawals after the incident. From 06:00 UTC on 1 November, withdrawals were restored in phases, starting with bitcoin, ether, USDT, USDC, BNB, XRP and others, with the most-requested networks first. By 12:36 UTC, all 15 networks and assets in phase one were open. As of 15:00 Tashkent time on 2 November, users could withdraw in 28 fiat currencies, including the US dollar, euro, Kazakhstani tenge and Kyrgyz som.
The numbers: between 06:00 and 15:30 UTC on 1 November, Binance processed $4.65 billion in withdrawals and $1.01 billion in deposits, for net outflows of $3.64 billion, or 2.1% of its $171 billion in total assets based on proof-of-reserves data as of 28 October. Independent trackers measure differently: Reuters, citing DefiLlama, reported $5.8 billion withdrawn, which Binance said included funds moved to its new hot wallet. So the sources disagree, and the gap comes down to methodology.
Binance says it will absorb the losses rather than pass them to users, and that it has covered about 95% of the affected assets; the remaining 5% are less liquid tokens. About $3 billion, mostly stablecoins, has been frozen, and Circle froze more than $1.1 billion in USDC.
Context
This is a crypto story, but at its core it is a classic payment-system failure: an automated payout channel, access to it through an internal system, and huge volume in a very short time. Billions leaving in 20 minutes suggests the weak point was not cryptography but the question of who, or which service, can issue a withdrawal instruction.
The second point is the scale of operational risk. Users are being made whole because Binance had the reserves (its SAFU fund and corporate funds) to do it. For a smaller exchange or payment company, an incident like this would mean insolvency.
It is relevant for Uzbekistan too. From 26 November, the Central Bank's rules for instant phone-number transfers, running 24/7 in real time, take effect. Speed is convenient for users, but errors and breaches travel at the same speed.
What it means
For users: a balance on a centralized exchange is a claim on that exchange. This time it was honoured, but that was a company decision, and while withdrawals were frozen there was no access to funds.
For businesses: an incident response plan has to exist before the incident: when to halt everything, how to tell customers, in what order to bring services back. Binance's phased reopening, prioritized by demand and announced publicly at each step, is a good template.
For backend and payments engineers:
- Separate signing from business logic. A service that can initiate a payout should not be able to sign it; an independent policy engine should check amount, destination and velocity.
- Velocity limits and circuit breakers. Put a hard cap on how much can leave a hot wallet per minute and per hour, with an automatic halt when it is exceeded. The same applies to banks: daily limits and automatic blocks on anomalous volume.
- A work laptop is not production. There should be no direct path from a DevOps machine to production containers: short-lived credentials, just-in-time access, two-person approval, a bastion and a full audit log.
- Keep the hot wallet small. Most funds belong in cold storage, with only operational float online.
- Ledger and reconciliation. To know your real net outflow, reconcile the internal ledger against external state continuously. The different tracker numbers here are exactly a methodology problem.
My take
For me, the main lesson is not "crypto is dangerous." It is that in any money system the automated payout channel is the most valuable asset, and no single internal service or single computer should be able to open it. The full attack details have not been disclosed yet, so a final verdict on Binance would be premature; the company itself has acknowledged "additional points of breach."
On the positive side, the crisis response was methodical: a fast halt, freezes in cooperation with stablecoin issuers, phased recovery and published numbers. Still, an architecture that let billions leave in 20 minutes raises real questions, and I'm waiting for the final technical report.
For banks and payment companies in Uzbekistan, especially teams connecting to instant payments and the national payment switch, this is a good moment to review their own kill switches, limits and the paths from work devices into production.
Sources: BleepingComputer, CNBC, TechCrunch, CoinDesk, Spot.uz, Wikipedia, Spot.uz — instant payments. This is personal opinion, not investment advice.
// comments
Loading comments...