Custody is the core of an exchange
Exchanges are usually judged by turnover, speed and market depth, but their most consequential job is custody: keeping customer assets safe and ensuring they move only with legitimate authority. A trading engine can be restored after an outage; assets sent on-chain cannot be recalled. Custody is therefore the foundation of customer trust and operational resilience, not a back-office service (Roy, 2025a, 2025b, 2026a).
What the Binance incident shows
On 29 October 2026 an attacker drained Binance hot wallets across nine to eleven blockchains, depending on the analyst. Binance estimates the impact at about US$6.3 billion; analytics firms put it between US$5.9 billion and US$6.43 billion, more than four times the US$1.46 billion taken from Bybit in 2025. Around US$3 billion has been frozen on-chain, and Binance says its SAFU fund and company resources will cover all losses (Binance, 2026a; Decrypt, 2026; Elliptic, 2026; TRM Labs, 2026).
How it happened matters more than how much. Binance first pointed to a compromised DevOps engineer’s laptop that gave access to a service container, from which unauthorised withdrawals were submitted, and reported no evidence of a private-key leak; a day later it said further points of breach had been found (Binance, 2026a, 2026b). On-chain evidence suggests the attacker did not steal keys but induced Binance’s own signing systems to authorise transfers. On Ethereum, an EIP-7702 delegation handed a hot wallet to an attacker-controlled contract that kept sweeping new inflows for hours after the initial drain (Elliptic, 2026). Attribution and the full attack chain remain provisional until Binance publishes its post-mortem. The amount stolen is not the opening hot-wallet balance, since funds arriving during the compromise were also exposed. And protecting private keys is not enough if a compromised system can obtain valid signatures.
Trading volume does not determine custody liquidity
Most trading on a centralised exchange settles through internal ledger entries; a trade between two customers need not touch the blockchain. Only external withdrawals, institutional delivery and some treasury movements do. Turnover is therefore a poor guide to the hot-wallet balance an exchange needs. The relevant measure is stressed external settlement demand, by token, network and time horizon, set against the time needed to replenish online liquidity safely.
Why hot–cold separation can fail
Hot wallets give immediate settlement but sit on online pathways; cold custody adds distance and slower, independent signing. The distinction holds only if a compromised online service cannot instruct or induce unrestricted transfers into or out of hot wallets. Automatic replenishment, broad delegated permissions, shared privileged identities, persistent smart-contract delegations and weak approval independence all erode it. Bybit showed the same from the other side: attackers compromised a developer’s machine at a third-party wallet provider and altered the transaction Bybit’s signers approved (Elliptic, 2026). In both cases the cryptography held; the systems and permissions around it did not. The common rule of thumb of keeping about 95 per cent of assets offline (Roy, 2025a; CoinDCX, n.d.) says little on its own about what can be moved under compromise.
Examining the Breach
Examining the breach would entail: wallet-level balances and movements, one to two years of withdrawals by asset and chain, replenishment logs, signing policies and limits, privileged-access records, and evidence of independent approval and emergency suspension. With these, a board, internal audit or an independent reviewer can reconstruct how much could leave in five minutes, an hour or a full replenishment cycle after a single-point compromise.
The regulatory backdrop
IOSCO’s 2023 recommendations cover custody and client-asset protection alongside operational and technological risk (IOSCO, 2023). The FSB’s July 2023 framework was strengthened specifically on safeguarding client assets (FSB, 2023), and Article 75 of the EU’s MiCA Regulation sets custody obligations, including liability for losses from incidents attributable to the provider (European Parliament and Council of the European Union, 2023).
The Binance incident suggests the next step in custody risk management is to combine transaction-demand evidence, custody architecture and worst-case transferability into one testable assessment of resilience.
- References
1. Binance. (2026a, October 30). Security incident update [Company announcement]. https://www.binance.com/en/support/announcement/detail/c5a8729a97ac486d963784d96c5642d7 - ]Binance [@binance]. (2026b, October 30). Investigation update: Our initial investigation suggested that the security incident may have involved a compromised laptop, however our subsequent investigation has uncovered additional points of breach [Post]. X. https://x.com/binance/status/2116089661408108712
3.CoinDCX. (n.d.). Fortifying the foundations: Wallet management and operational resilience in crypto exchanges [Policy blog on the work of Dr. Sunando Roy]. https://coindcx.com/policy/fortifying-the-foundations-wallet-management-and-operational-resilience-in-crypto-exchanges/\ - Decrypt. (2026, October 30). Binance finds ‘additional points of breach’ in $6.3B hack. https://decrypt.co/382203/binance-finds-additional-points-of-breach-in-6-3b-hack
- Elliptic. (2026, October 30). The largest theft in history: North Korea likely behind the $6.4 billion Binance hack. https://www.elliptic.co/insights/binance-largest-hack-in-history/
- European Parliament and Council of the European Union. (2023). Regulation (EU) 2023/1114 of 31 May 2023 on markets in crypto-assets. Official Journal of the European Union, L 150. https://eur-lex.europa.eu/eli/reg/2023/1114/oj
- Financial Stability Board. (2023, July 17). FSB global regulatory framework for crypto-asset activities. https://www.fsb.org/2023/07/fsb-global-regulatory-framework-for-crypto-asset-activities/
- International Organization of Securities Commissions. (2023, November 16). Policy recommendations for crypto and digital asset markets: Final report (FR11/2023). https://www.iosco.org/library/pubdocs/pdf/IOSCOPD747.pdf
- Roy, Sunando. (2025a). Wallet management practices in cryptocurrency exchanges: Security, compliance, and future innovations. SSRN. https://ssrn.com/abstract=5124001
- Roy, Sunando. (2025b). Operational resilience in crypto exchanges: Strengthening controls for a secure digital asset ecosystem. SSRN. https://doi.org/10.2139/ssrn.5124017
- Roy, Sunando. (2026a, February 11). Risk management in crypto asset service: Balancing innovation with operational resilience. Sunando Roy – On Banking, Finance and Society. https://sunandoroy.org/2026/02/11/risk-management-in-crypto-asset-service-balancing-innovation-with-operational-resilience/
- Roy, Sunando. (2026b, February 13). Proof of reserves: Not the silver bullet for crypto transparency. Sunando Roy – On Banking, Finance and Society. https://sunandoroy.org/2026/02/13/proof-of-reserves-not-the-silver-bullet-for-crypto-transparency/
- TRM Labs. (2026, October 29). North Korea steals USD 6.36 billion from Binance in the largest hack in history. https://www.trmlabs.com/resources/blog/north-korea-steals-usd-6-36-billion-from-binance-in-the-largest-hack-in-history
Annexure: What experts are saying about the Binance incident
1. Scale of the theft
Elliptic puts the theft at US$6.43 billion across 11 blockchains and, in its own assessment, calls it almost certainly the largest theft of any kind (Elliptic, 2026). TRM Labs estimates US$6.36 billion across ten chains, most of it within about 20 minutes (TRM Labs, 2026). Arkham counts US$6.01 billion on nine chains (Decrypt, 2026), while Eric Jardine of Chainalysis put the minimum at US$5.9 billion and stressed that the cause was unconfirmed (Reuters, 2026). Nicholas Turnbull (Crypto Consulting Australia) called it clearly the largest crypto hack to date (ABC News, 2026). The differing figures and chain counts reflect each firm’s methodology and are best read as source-specific.
2. Who is behind it
TRM Labs attributes the attack with high confidence to TraderTraitor, a North Korea-linked group, citing long preparation and signatures obtained in advance (TRM Labs, 2026); Jonathan Levin of Chainalysis reached the same view (CNBC, 2026). Ido Ben-Natan (Blockaid) and investigator ZachXBT also pointed to North Korea, the latter naming the Lazarus Group as the likely attacker (Sandor & Alpher, 2026). Elliptic judges North Korean involvement likely but cautions that the on-chain overlap with the Bitget exploit is small, the laundering of stablecoins was unusually slow, and no government has confirmed attribution (Elliptic, 2026).
3. How the attack worked, and what it says about custody
Binance first pointed to a compromised engineer’s laptop, then reported additional points of breach; chief executive Richard Teng said the attack was not a single one-click action (Binance, 2026b; Decrypt, 2026). Elliptic found, on the evidence available on 30 October, that Binance’s own systems were induced to sign transactions, including through an EIP-7702 delegation that kept sweeping deposits; as with Bybit, the people and infrastructure around the wallets failed, not the cryptography (Elliptic, 2026). Jeff Ko (ViaBTC) likewise called it a signing failure rather than a smart-contract exploit (Reuters, 2026).
Mark Karpeles (former head of Mt. Gox) suggested the attacker gained valid signing authority over two addresses, possibly via a private key, and that spreading assets across addresses limited the damage (Reuters, 2026). This conflicts with Binance’s statement that no key leaked (Binance, 2026a); both positions remain provisional. Sean Foley (Macquarie University) noted that assets on an exchange are only as safe as the exchange, and that bigger holdings make a bigger target (Al Jazeera, 2026).
4. Solvency, liquidity and markets
Sean Farrell (Fundstrat) saw no meaningful solvency risk for Binance and named the pace of withdrawal resumption and the orderliness of outflows as the indicators to watch (CNBC, 2026); this is an analyst’s opinion, not a solvency assessment. Tim Sun (HashKey) flagged liquidity disruption across venues; Aravind Srinivasan (GSR) saw spot bitcoin spreads at about five times normal; Damon Xu (1Token) expected less cross-exchange trading and leverage; and Geoffrey Kendrick (Standard Chartered) expected further price falls until frozen assets are released and details emerge. Yuya Hasegawa (bitbank) and Kenneth Goh (UOB Kay Hian) saw the main risk as lost confidence in centralised exchanges, though Hasegawa thought it could fade if service resumes quickly (Reuters, 2026).
5. Recovery and the policy response
Early figures on 29 October showed Tether freezing more than US$1.5 billion of USDT and Circle about US$800 million of USDC (Sandor & Alpher, 2026). Elliptic estimated around US$3.33 billion of native assets, which no issuer can freeze, still under the attacker’s control at about 07:00 UTC on 30 October; these are most readily intercepted at exchanges or other intermediaries, though legal process may also apply (Elliptic, 2026). Joshua Chu (Hong Kong Web3 Association) framed the aftermath across regulation, law enforcement, civil recovery and preventing exchanges from moving stolen funds, and criticised industry reluctance to work with governments (Reuters, 2026).
Overall
1. TRM Labs and Elliptic locate the failure in the transaction-authorisation infrastructure, which supports a focus on the whole authorisation chain rather than key protection alone; Binance’s disclosure of further breach points means the exact mechanism is still unresolved.
2.Second, transfers continued after the initial drain, so exposure must be measured over time, as MCAE proposes. Third, the theft’s liquidity and confidence effects can impair markets even when an exchange promises to make customers whole. 3. All of this is preliminary until independent forensic findings and Binance’s post-mortem are published.



Leave a Reply