On Oct. 29, attackers stole roughly $6.3 billion in digital assets from the world's largest cryptocurrency exchange, Binance. The FBI attributed the attack to North Korea.
What is interesting is how the attackers did what they did. The path in wasn’t new: untrusted software, a convincing recruiting scheme, compromised endpoints, an internal AI tool and access that could not be contained quickly enough.
These issues crossed multiple departments, with everyone assuming someone else had it covered. The psychological term for this is the bystander effect. That's often where security breaks down, and the CIO ends up having to help hold it all together.
Binance had money, a dedicated security team and advance warnings about this exact tactic, and it still succumbed to those weaknesses. The weaknesses are quite common, so CIOs and other IT leaders should understand and learn from Binance's errors.
What happened with Binance and why it matters
The campaign against Binance unfolded over several months, according to the company. In June, a Wallet DevOps engineer who was attempting to install an AI coding tool downloaded a tampered Homebrew installer. Security replaced his laptop and rotated the related credentials, but the attackers had already learned about his role and the people he worked with.
In September, a fake recruiter contacted a different engineer through a legitimate job search site. After interviews with a phony HR representative, he participated in an interview with a supposed CEO and, during that interview, downloaded an assignment onto his Mac work laptop that included a malicious code repository. The malware evaded endpoint detection.
On Oct. 20, the attackers used that machine and an internal AI tool to reach two wallet engineers' computers, one of which had a token with container deployment permissions. On Oct. 29, they used the stolen deployment permission token to run a malicious script against Binance's wallet-signing infrastructure.
The damage reached well beyond the stolen funds. Binance suspended withdrawals, and customers couldn't move their money until withdrawals became available again in phases on Nov. 1. Binance eventually experienced $3.64 billion in net outflows.
The attack on Binance matters to CIOs and other technology executives because none of the entry points were unique to cryptocurrency. A recruiter's message, a developer laptop, an internal AI tool and a long-lived token exist in nearly every organization. The CIO is usually the one who must see how it all connects and keep IT from falling apart when things go sideways.
The importance of privileged access and identity governance
The attackers didn't have to break into Binance's wallet systems directly. They borrowed access that already existed. Every step relied on access someone had legitimately granted.
Identity governance can prevent this outcome. Inventory every identity that can touch production or money-moving systems, including service accounts, API keys, deployment tokens and AI agents. Give each one an owner and review its access against what it actually does day-to-day.
Remove standing privilege wherever possible. Just-in-time access helps prevent longstanding permissions because it involves, for example, an engineer requesting deployment rights for a specific task, someone approving it and then the access expires. Then require a second approver for systems that move money and require administrators to work from dedicated, locked-down workstations.
AI tools demand the same scrutiny. Binance later rebuilt the internal AI tool that was used in the attack and its access controls. When organizations create AI policies, they often overlook the integrations and permissions that make AI tools useful and, if compromised, dangerous.
Treat AI assistants and agents that can access internal data or carry out actions like any other sensitive identity. Find out what they can read, change, send and retain.
Taking action against malware
Malware opened both doors at Binance. Its report calls the malware a zero-day, but it does not establish that the attackers exploited a previously unknown vulnerability in macOS. Macs are not immune to malicious code, especially when people run it themselves.
Binance was clearly aware of the dangers of fake job interviews. Binance chief security officer Jimmy Su described the fake interview coding assignment tactic, which closely resembles a prior North Korea-linked campaign documented by security researchers and government agencies, in a July interview with Cointelegraph. He also said Binance ran monthly phishing simulations for years. Awareness helps, but it does not prevent an endpoint from executing untrusted code.
Employee training still matters, but it needs to be specific. Show technical staff exactly how these recruiting scams work, from the first message to the take-home test. HR has a role to play and can help communicate with employees.
Make it clear that coding assignments and installers from recruiters must never run on company machines, and create an isolated environment for legitimate testing. Encourage employees to report any problems early so they won’t hide incidents out of fear.
Also, keep up with updates and patches, including for Macs and developer tools, which are often not part of the overall patching cycle. Still, nothing in Binance's account suggests that patching would have stopped this attack, as the engineer ran the code themselves. Developer workstations are often among the most powerful and least controlled machines.
A blanket ban on all third-party code is not realistic. What works is approved software sources, removal of unnecessary local admin rights and tightly scoped credentials with limited lifetimes. It’s also important to use disposable sandboxes for anything untrusted and include those systems in ongoing penetration testing.
Don’t forget third-party visibility
Binance's problem became a lot of other people's problem within hours. Businesses and traders that depend on the exchange were cut off from their funds for days. When a cloud provider, payment processor or SaaS platform that a company relies on gets hit, that company inherits the downtime and sometimes the attacker, too. CIOs must determine how their company would deal with that situation.
Know which partners the company relies on for operations and what to do if one goes dark for multiple days. Map every connection those partners have to the company environment, including API keys, OAuth grants and shared credentials, and revoke or rotate them as soon as one reports a breach. Expect attackers will send phishing attempts impersonating the breached company. Adding breach notifications and cooperation requirements into partner contracts before they’re needed can be helpful later.
The Binance hack also led to examples of good partner behavior. Tether and Circle began freezing stolen stablecoins quickly after the attack, and Circle froze $823 million in stolen USD coin, a type of stablecoin, before the attackers could swap it.
Overall, when a partner gets hit, help them contain the damage. Set up security contacts before anything happens, share indicators of compromise and don't cut off partners by reflex when a temporary restriction will work. Any company could be next and will want the same courtesy.
Plan containment measures now
Companies must prepare a containment strategy before an attack happens.
Attackers could use an OAuth grant, an API key or a session that survives a password reset. Containing a compromised engineer's access is similar to offboarding an employee but comes with more pressure. Decide in advance what is revoked and who can authorize it.
Motivational speaker Brian Tracy wrote in his book “The 100 Absolutely Unbreakable Laws of Business Success,” "Every minute spent in planning saves ten minutes in execution." That’s a great philosophy for IT and incident response.
Common mistakes as seen in the Binance hack
In the context of the Binance incident, common cybersecurity mistakes include the following:
• Treating employee awareness as the final control. Training helps, but technical safeguards must still hold when someone is deceived.
• Giving privileged developer machines a pass. The more a workstation can reach, the more scrutiny it deserves.
• Putting too much faith in endpoint detection and response/extended detection and response. Those controls are valuable, but attackers can find ways around them.
• Publishing an AI policy without mapping access. Integrations and service identities that haven’t been reviewed pose risks that a policy alone won't resolve.
• Resetting passwords while overlooking tokens and delegated permissions. Containment must address every authorization path.
• Failing to bring in HR, legal and the communications department until it’s time to write the press release. These people must be involved long before an incident becomes public.
3 things CIOs should do before the end of the year
An incident this large will lead to various product pitches, resolutions and promises. CIOs should carry out these three tasks before the year is over:
1. Isolate untrusted code from production. Publish clear expectations with HR, provide disposable sandboxes, enforce the use of approved software sources only and keep unnecessary production credentials off general-purpose developer workstations.
2. Map AI integrations and privileges. Inventory internal AI tools, their owners, connected data, permitted actions and service identities. Limit permissions, and include these tools in security testing to verify their connections can be disabled quickly.
3. Test containment from start to finish. Run the Binance scenario with IT, security, HR, legal, risk and communications. Include compromised developer endpoints, deployment tokens, active SaaS sessions and authorizations that can survive ordinary shutdowns. Set a target containment time and see whether the team can meet it.
The endgame at Binance depended on cryptocurrency technology, but getting inside did not. A fake recruiter, untrusted software, a compromised workstation and poorly understood permissions can show up in any organization.
Know where those paths lead, test whether controls stop them and make sure the right people have the authority to act. Binance's security chief warned about one of the tactics used against them months before, but simply knowing about it wasn't enough.
Kevin Beaver is an independent information security consultant, writer and professional speaker with Atlanta-based Principle Logic, LLC. With more than 30 years of experience in the industry, Beaver specializes in performing vulnerability and penetration tests as well as virtual CISO consulting work.