The theft of approximately $351.6 million from cryptocurrency exchange Bitget is another major reminder that securing digital assets requires far more than protecting private keys. According to Bitget, attackers compromised a critical backend system within its wallet infrastructure, manipulated transaction data, and caused the platform’s authorization process to approve transfers of funds from hot and warm wallets. Bitget says its cold wallets were unaffected and that the attackers did not obtain the private keys of either its cold or hot-wallet infrastructure.

The breach was detected at approximately 18:31 UTC on September 24, 2026, when Bitget’s monitoring systems identified unauthorized transfers involving a limited number of hot wallets. The company estimates that around $351.6 million worth of assets were affected. The stolen assets reportedly included ETH, XRP, BNB, AVAX, USDT, and USDC across multiple blockchain networks including Ethereum, XRP Ledger, Arbitrum, Avalanche, Optimism, BNB Smart Chain, and Base. Withdrawals were subsequently suspended while the exchange initiated a wider security review.

The technical description provided by Bitget is particularly important. CEO Gracy Chen said the attackers compromised a backend system associated with the exchange’s wallet service and exploited it to forge transaction information that was then passed into Bitget’s signing and authorization process. In other words, the attackers may not have needed direct possession of the private keys if they could manipulate the trusted system responsible for deciding what should be signed.

This distinction fundamentally changes how the incident should be viewed. A private key can be strongly protected inside hardened infrastructure, hardware security modules, or tightly controlled signing systems, but those protections provide limited value if an attacker compromises the upstream system that tells the signing infrastructure which transaction is legitimate. The cryptography may still work perfectly. The problem is that it may be cryptographically approving attacker-controlled information.

That makes the Bitget incident a useful example of the difference between key security and transaction integrity. Protecting private keys is essential, but organizations must also ensure that every system feeding information into the signing process is trustworthy. The transaction request, destination address, amount, approval context, user identity, risk score, policy decision, and signing command all form part of the security boundary.

The attack flow described so far can be summarized as: backend compromise → manipulated transaction data → legitimate authorization workflow invoked → valid signatures generated → cryptocurrency transferred to attacker-controlled wallets. If Bitget’s preliminary description is confirmed by the ongoing investigation, this means the attackers did not defeat the cryptographic signing mechanism directly. They manipulated the system around it.

This type of attack is particularly dangerous in cryptocurrency infrastructure because valid blockchain transactions are generally irreversible. Once the signing system approves and broadcasts a transaction, there is no traditional financial clearing process capable of simply reversing it. Blockchain foundations, stablecoin issuers, exchanges, and analytics firms may sometimes help identify, freeze, or intercept stolen assets, but recovery becomes a race against the attacker’s ability to move and launder those funds.

Bitget says it contacted organizations associated with the affected blockchain networks and that some wallet addresses linked to the attackers were frozen. The company also reported that at least some funds had been recovered, although the amount had not been publicly specified at the time of initial reporting.

The company has brought in Mandiant and SlowMist to assist with the investigation. That is significant because the most important unanswered question is still the initial intrusion vector. Bitget has confirmed the compromise of a backend wallet-related system, but investigators were still determining exactly how the attackers obtained access to that environment.

Without that information, speculation about whether the breach began through phishing, stolen credentials, supply-chain compromise, exposed infrastructure, insider access, or exploitation of a vulnerability would be premature. The immediate technical evidence establishes the consequence of the intrusion more clearly than its origin.

Suspected North Korean involvement should remain an assessment, not a certainty

Bitget has said the attack shows similarities to activity previously associated with North Korean threat actors. According to Gracy Chen, investigators identified IP addresses that appeared to use VPN infrastructure matching patterns seen in previous DPRK-linked attacks. Bitget also said the transaction behavior and on-chain movement showed similarities to past North Korean cryptocurrency operations.

However, attribution remains preliminary.

The available indicators may support suspicion, but VPN usage, infrastructure overlap, and cryptocurrency-laundering behavior are not by themselves definitive proof of state attribution. The investigation should therefore continue to describe North Korean involvement as suspected or assessed, rather than established fact, unless stronger technical or intelligence evidence emerges.

That caution is especially important because cryptocurrency attacks attributed to North Korean groups have become so frequent that there is a risk of treating any large crypto theft as automatically DPRK-linked. Attribution should remain evidence-driven rather than reputation-driven.

At the same time, the hypothesis is plausible in the broader threat landscape. North Korea-linked groups have repeatedly targeted cryptocurrency exchanges, blockchain developers, DeFi platforms, and Web3 organizations. A multinational advisory released only days before the Bitget incident described the long-running Contagious Interview campaign, in which North Korean actors allegedly compromised more than 30,000 devices and targeted thousands of cryptocurrency wallets through fake recruitment and coding exercises.

North Korea-linked TraderTraitor activity has also previously been associated with very large cryptocurrency thefts, including the 2025 Bybit breach, which resulted in approximately $1.5 billion in losses and was publicly attributed to North Korea by the FBI. This history makes exchanges and wallet infrastructure especially high-value targets because successful compromises can generate enormous amounts of cryptocurrency that can potentially be moved across borders rapidly.

Hot wallets remain the operational weak point

Bitget says the breach affected part of its hot and warm wallet infrastructure while its cold wallets remained secure. This reflects one of the unavoidable tradeoffs in exchange architecture. Cold wallets can provide stronger protection because their keys or signing processes remain isolated, but exchanges still require hot wallets to support customer withdrawals, liquidity, market operations, and other real-time activity.

The operational requirement to keep some assets readily available creates a permanent attack surface.

Attackers therefore do not necessarily need to compromise the exchange’s entire asset reserve. Gaining control over a sufficiently funded hot or warm wallet system can still produce hundreds of millions of dollars in losses.

The Bitget incident demonstrates why wallet security needs several independent control layers. A transaction should ideally require not only a valid cryptographic signature but also independent validation of destination addresses, amounts, transaction velocity, historical behavior, approval context, and aggregate exposure.

Signing infrastructure should distrust the backend that calls it

One of the strongest architectural lessons from this incident is that a signing system should not automatically trust transaction data simply because it originates from an internal backend.

Internal systems can be compromised.

A more resilient design treats even internal requests as potentially hostile and independently verifies whether they conform to expected business rules.

That can include controls such as destination allowlisting, transaction limits, withdrawal velocity thresholds, independent risk engines, multi-party approval, delayed settlement for unusual transfers, behavioral anomaly detection, and separate verification paths before high-value transactions are signed.

For exceptionally large transfers, an authorization request originating from a backend service should ideally require additional validation from a system that the backend cannot modify.

This follows a basic security principle: the system creating a transaction should not also be the only system responsible for deciding whether that transaction is trustworthy.

Transaction signing needs contextual integrity

Many cryptocurrency security designs focus heavily on the protection of keys. That is necessary, but insufficient.

A signer should ideally verify not only:

“Is this request cryptographically valid?”

but also:

“Does this transaction make sense in the context of the business operation that supposedly created it?”

For example, an exchange signing system might independently evaluate whether a withdrawal corresponds to an actual user request, whether the destination address was correctly registered, whether the amount fits the customer’s balance, whether approvals match the transaction being signed, and whether an unusual concentration of withdrawals is occurring.

This type of contextual validation can make attacks against upstream backend systems considerably harder to convert into successful asset theft.

Segmentation between wallet components matters

The fact that Bitget says its cold wallets were unaffected demonstrates the value of segmentation. A compromise of one operational layer should not automatically provide control over every wallet tier.

The same principle should apply within the hot-wallet environment itself.

Services responsible for receiving withdrawal requests, risk scoring, transaction construction, authorization, signing, and broadcast should ideally be separated both logically and operationally. Credentials should be scoped narrowly, and one compromised backend service should not be able to independently fabricate an entire high-value transaction lifecycle.

The greater the financial exposure of the wallet, the stronger that separation should become.

Monitoring should focus on transaction behavior as well as infrastructure

The incident was discovered because Bitget detected unusual transfers. That reinforces the value of real-time transaction analytics in cryptocurrency environments.

Infrastructure monitoring alone may detect malware, unauthorized logins, or suspicious processes, but transaction-level monitoring may reveal the financial consequence more quickly.

Security teams should therefore watch for patterns including sudden increases in withdrawal volume, unusual destination addresses, abnormal cross-chain activity, transfers inconsistent with customer behavior, repeated large transactions, unexpected interaction with bridges or decentralized exchanges, and transaction requests that bypass normal workflow patterns.

An attacker who successfully compromises backend infrastructure may attempt to make every server-side action look legitimate. The blockchain activity itself can still reveal that something unusual is happening.

Asset movements should trigger independent kill mechanisms

High-value cryptocurrency platforms should consider independent mechanisms capable of stopping withdrawals when transaction patterns exceed expected thresholds.

Such mechanisms should be difficult for the same compromised backend system to override.

A separate monitoring platform could, for example, suspend signing when aggregate withdrawals exceed a defined limit, when funds suddenly move to newly observed addresses, or when transactions appear across multiple chains in an unusual pattern.

The objective is to create controls where compromising one system is not enough.

Attackers should have to defeat several independent decision points before hundreds of millions of dollars can leave the organization.

Protection funds help customers, but do not reduce the security significance

Bitget says the entire incident falls within its User Protection Fund, which held more than $464 million at the time of the incident, and that user account balances remain protected.

That is important from a customer-impact perspective.

However, insurance or protection reserves should not be confused with security controls.

A fund can absorb financial losses after a breach. It cannot prevent the breach, protect stolen operational secrets, or guarantee that attackers did not establish deeper persistence within the environment.

The availability of sufficient reserves therefore limits customer financial impact, but it does not diminish the seriousness of a $351.6 million backend compromise.

The incident should trigger compromise assessment beyond wallet systems

If attackers gained access to a critical wallet backend, Bitget and similar organizations investigating comparable incidents need to determine what other systems were reachable from that environment.

Questions should include whether attackers accessed source-code repositories, CI/CD systems, deployment infrastructure, databases, cloud-management consoles, identity providers, monitoring platforms, or secrets used by other internal services.

The attacker’s objective may have been cryptocurrency theft, but that does not guarantee the attacker restricted themselves to systems necessary for the theft.

A financial objective can coexist with extensive reconnaissance and persistence.

Credential rotation should be broad enough to address lateral movement

If the compromised system contained credentials, tokens, certificates, SSH keys, API credentials, or service-account secrets, those should be considered potentially exposed.

Remediation should therefore include more than rebuilding the affected backend.

Security teams should identify credentials accessible from the compromised environment, rotate them, review their historical use, and determine whether they were used elsewhere.

Otherwise, the attacker may lose access to the wallet system while retaining access to another internal platform.

Cryptocurrency platforms remain exceptionally attractive targets

The economics make exchanges unusually attractive to sophisticated attackers.

A conventional enterprise breach may require months of extortion, negotiation, or stolen-data monetization.

A successful cryptocurrency infrastructure compromise can potentially move hundreds of millions of dollars within minutes.

That creates enormous incentives for attackers to invest significant time in reconnaissance, employee targeting, supply-chain compromise, custom malware, cloud infrastructure attacks, and social engineering.

For organizations operating cryptocurrency infrastructure, the security standard therefore has to reflect the value directly accessible through their systems.

The broader cybersecurity lesson

The Bitget incident illustrates why cryptocurrency security cannot stop at protecting private keys.

If an attacker compromises the workflow that decides what the private key should sign, then the cryptography may become an unwilling participant in the theft.

The security chain therefore needs to protect:

request generation → identity verification → risk assessment → transaction construction → authorization → signing → broadcast → post-transaction monitoring

Each stage should be independently validated wherever practical.

If Bitget’s preliminary findings are confirmed, the central lesson from the attack is not that private-key protection failed.

It is that the trusted software surrounding those keys was manipulated.

That is arguably more important for the wider cryptocurrency industry because many organizations have invested heavily in securing cryptographic material while continuing to treat internal backend systems as trusted.

They should not.

A hardware-protected private key can still sign a fraudulent transaction if the system feeding it fraudulent information has been compromised.

The safest architecture therefore assumes that even internal wallet infrastructure can eventually be breached and designs the transaction-authorization process so that compromising one component is never enough to move hundreds of millions of dollars.

The Bitget breach is a reminder that in cryptocurrency security, protecting the key is only half the problem.

You must also protect the truth of what the key is being asked to sign.

Cryptocurrency exchange Bitget said suspected North Korean threat actors have stolen $351.6 million from its hot and warm wallets. "At 18:31 UTC on September 24, 2026, Bitget's security systems identified unauthorized transfers involving a limited number of hot wallets," BitGet said in a post shared on X. "Bitget's cold wallets and the overwhelming majority of platform assets remain

Source: Bitget Says Suspected North Korean Hackers Stole $351.6M After Backend Compromise via The Hacker News — published 25 Sep 2026.