Updated on: November 12, 2026
Crypto's biggest hacks have exposed a recurring weakness: attackers don't always need to break the code to get to the money.
In several of the largest incidents, they gained access through compromised employees, third-party software, stolen credentials, or transaction-signing systems. Six of the ten largest thefts involved failures in these areas, accounting for approximately 84% of the combined value stolen.
Code vulnerabilities tell a different part of the story. They caused the other four: smart contract bugs at Poly Network and Wormhole and proof verification flaws in the BNB Bridge and Liquid Network software. Code bugs also remain a recurring source of security breaches across the industry.
Looking at these incidents together reveals two distinct challenges for crypto organizations: protecting the systems and people that authorize transactions and ensuring that the code governing digital assets behaves as intended.
The 10 biggest crypto hacks in history
What caused the biggest crypto losses
- BinanceSocial engineering to the signing service$6.34B
- BybitCompromised signing interface$1.50B
- Ronin BridgeValidator key compromise$624M
- CoincheckHot wallet custody$530M
- BitgetThird-party software to the signing service$387.5M
- DMM BitcoinWallet vendor compromise$308M
- Poly NetworkUnprotected privileged function$611M
- BNB BridgeProof verification bug$586M
- WormholeSignature verification bug$320M
- Liquid NetworkProof verification bug$319M
How each of the biggest crypto hacks happened
Binance hack: $6.34B, October 2026
What did the attacker do: For four months, the attacker worked on Binance engineers: a tampered installer in June, a fake job interview in September and an internal AI tool in October. With a stolen deployment token, the attacker launched a container that asked Binance's wallet signing service to sign transfers outside the normal withdrawal process. The service signed them on every chain, so no private key had to be stolen. North Korea's TraderTraitor group carried out the theft.
Why the system allowed it: The signing service accepted requests from inside Binance's infrastructure without checking them against an approved withdrawal, so one engineer's stolen token was enough to move funds.
Our Binance hack analysis walks through each step.
Bybit hack: $1.5B, February 2025
What did the attacker do: The attacker compromised a Safe{Wallet} developer's machine and planted malicious JavaScript in the wallet interface two days before the theft. When Bybit's signers approved a routine transfer from a cold to a hot wallet, their screens showed a normal transaction. The data they actually signed handed control of the cold wallet to an attacker-controlled contract, which emptied it.
Why the system allowed it: Bybit's multisig required several approvals, but every signer relied on the same web interface, and nothing independent checked what they were signing. Bybit covered the loss and kept withdrawals open.
Hacken's investigation walks through the injected code, and tracing the Bybit billion follows the money afterward.
Bitget hack: $387.5M, September 2026
What did the attacker do: The attackers got in through flaws in two third-party security products and reached a wallet server. A purpose-built withdrawal tool then forged the risk-control fields on withdrawal requests, and Bitget's own signing service signed the transfers.
Why the system allowed it: Bitget detected the theft quickly, but detection couldn't stop the signer. The reconciliation alert fired about six minutes after the first large transfer, and the signing service kept running for another 2 hours 39 minutes. About $238M of the $387.5M left in that window.
DMM Bitcoin hack: $308M, May 2024
What did the attacker do: A North Korean actor posing as a LinkedIn recruiter sent an employee of Ginco, DMM Bitcoin's wallet software vendor, a Python script disguised as a pre-employment test. Session cookies stolen from that machine let the attacker impersonate the employee and manipulate a legitimate DMM Bitcoin transaction request, taking 4,502.9 BTC.
Why the system allowed it: The change came through a trusted vendor's account, so it looked like routine work to DMM Bitcoin's own controls.
Ronin Bridge hack: $624M, March 2022
What did the attacker do: A fake job offer sent to a Sky Mavis engineer gave the attacker four of Ronin's nine validator keys, one short of the five needed to approve a withdrawal. The fifth came from an old shortcut: Axie DAO had let Sky Mavis sign on its behalf during a traffic surge in late 2021, and that permission was never revoked. With five signatures, the attacker withdrew 173,600 ETH and 25.5M USDC. Nobody noticed for six days, until a user couldn't withdraw.
Why the system allowed it: A five-of-nine threshold looked strong, but four keys sat with one company and a temporary permission outlived its purpose.
Coincheck hack: $530M, January 2018
What did the attacker do: At 2:57 a.m. Tokyo time on January 26, 2018, an attacker started draining Coincheck's NEM hot wallet. Several employees had opened emails carrying malware, and from there the attacker reached the wallet's private key. In eight transactions over about 20 minutes, 523 million NEM belonging to roughly 260,000 customers went to the attacker's address, and more followed later that morning. Coincheck noticed the missing balance eight and a half hours after the break-in.
Why the system allowed it: One private key controlled the entire balance. Coincheck kept its customers' NEM in an internet-connected wallet with no multisignature, even though it used multisig for bitcoin and kept most of its bitcoin and ether offline. Its president blamed technical difficulty and a shortage of staff. The exchange was also trading without a license: Japan let existing exchanges keep operating while their registration was pending, and Coincheck's application had been waiting since September 2017.
Poly Network, BNB Bridge, Wormhole and Liquid Network: $1.84B lost to forged messages and proofs, 2021–2026
Bridges and cross-chain protocols release or mint assets when they receive a valid message or proof from another chain. In these four cases, the code that checked validity had a flaw, so the attacker could fake a message or proof and take assets that were never deposited.
- Poly Network lost $611M because the contract that handles cross-chain messages could be made to call a privileged function, which let the attacker replace the keys that approve withdrawals. The attacker returned the funds.
- BNB Bridge lost $586M after the attacker forged a withdrawal proof that passed a flawed tree verification.
- Wormhole lost $320M when its Solana contract accepted a spoofed system account in place of the guardian signature check, and the attacker minted 120,000 wETH with no ETH behind them.
- Liquid Network lost $319M when a flaw in how Elements, Liquid's validation software, caches range-proof checks let the attacker mint about 4,000 unbacked L-BTC and redeem them for real bitcoin through the federation. The attacker returned ~85% after Blockstream patched the bug.
We released our guide on cross-chain bridge security so anyone can check basics: who verifies messages, how many verifiers must agree, and what happens when one fails.
Crypto hack trends: how the causes change from year to year
Access control failures. Compromised employees, third-party software, credentials, and transaction-signing infrastructure account for some of the largest losses in crypto. Their financial impact is disproportionate because a single successful intrusion can expose assets across multiple wallets or blockchains.
The share of losses attributed to access control varies significantly between reporting periods. A single large exchange breach can outweigh dozens of smaller incidents, making quarterly percentages particularly sensitive to individual events.
The Binance, Bybit, Ronin, Bitget, and DMM Bitcoin incidents illustrate different routes to the same outcome: attackers gained access to systems or credentials that could authorize transactions.
How the causes of crypto losses changed, 2025 to Q3 2026
- 2025$4.0B lost53%$2.12B24%$952M13%$512M
- Q1 2026$482.7M lost15%$71.9M63%$306M18%$86.2M
- Q2 2026$764.0M lost88%$674.9M
- Q3 2026$1.18B lost40%$477.6M48%$569.5M
The largest thefts are long operations. The attacker works on engineers, vendors and infrastructure for months and needs only one path to keys, signing systems or custody.
Old bug classes come back. Some crypto incidents involve vulnerability classes that have been understood for years. Integer overflows, flawed verification logic, and outdated contracts continue to appear in production environments.
The Truebit exploit in 2026 demonstrated how an old integer overflow vulnerability could still affect a deployed contract. Other incidents involved deprecated contracts that remained callable even after their functionality had been removed from the user interface.
Removing a contract from an interface does not remove it from the blockchain. Organizations need to account for deployed code, permissions, and integrations throughout the lifecycle of a protocol, not just during its initial release.
The oldest custody rule broke as well. Bitget's and Binance's signing services approved withdrawals they couldn't tie to a legitimate request. A CCSS audit tests that rule.
How to protect crypto organizations from hacks
The incidents in this ranking involved different technical weaknesses, but they point to a common requirement: security assessments must cover more than smart contract code.
A smart contract audit can identify vulnerabilities in transaction logic, access controls, and cross-chain verification. It cannot establish whether an employee's device is compromised, a third-party vendor has been breached, or a signing service will approve an unauthorized request.
Addressing these risks requires several complementary security measures:
- Smart contract audits to identify vulnerabilities in deployed code, privileged functions, and transaction verification.
- Penetration testing to assess exposed applications, APIs, cloud environments, and wallet infrastructure.
- Red teaming to simulate attack paths involving employees, suppliers, developer environments, and transaction-signing systems.
- Custody and key management assessments to evaluate access restrictions, signing policies, transaction authorization, and recovery procedures.
- Incident response exercises to establish whether security teams can detect suspicious activity and stop unauthorized transfers before losses escalate.
Hacken's red team runs these attack paths end to end, from social engineering and developer machines to CI/CD, containers, and wallet and key management. Findings are mapped to MITRE ATT&CK, engagements follow the DORA threat-led penetration testing methodology, and purple team sessions time your detection and response against the live attack. For a narrower scope, a penetration test covers web, API, mobile, cloud and wallet infrastructure, with a free retest after fixes. Our exchange security assessment methodology lists what an assessor reviews.





