Skip to main content
Return to TrendAI 保安網誌

Two Breaches, One Lesson: Why Your Security Program Can’t Rely on Trust

A record $6.3 billion crypto theft and a university-wide shutdown hit within two weeks of each other. Different victims, same weakness: trust that was granted once and never re-examined. Here’s what both incidents show, and four controls to check this month.

Perspectives General Markets

The biggest crypto theft in history and a university shutdown happened within two weeks of each other. They look unrelated: one involved a cryptocurrency exchange, the other a Nashville campus. But both point to the same weaknesses that exist in most organizations, and both are worth a close look from anyone responsible for security.

What happened at Binance

On October 29, roughly $6.3 to $6.4 billion moved out of Binance wallets across ten blockchains, most of it in about 20 minutes. The FBI has attributed the attack to TraderTraitor, a North Korea-aligned hacking group. Binance’s own incident report describes a months-long campaign, not a single breach. In June, a wallet engineer downloaded a tampered installer for a developer tool found through a search engine. Binance caught it, replaced the device and reset the engineer’s credentials, but information about the engineer and colleagues had already reached the attackers. A second engineer was later compromised through a fake job interview, in which a coding assignment installed custom malware that endpoint protection did not detect. From there, the attackers moved through the network, abused an internal AI tool, and reached the service that signs wallet transactions.

What stands out is how the attack was prepared. According to blockchain analysts , the attackers funded their wallet through a cryptocurrency mixer, then submitted 50 pre-signed authorizations in a single transaction, each for a different transaction counter, so that one would match and hand them control of an Ethereum hot wallet. The theft didn’t need to break encryption or defeat a firewall. It needed to exploit trust that had already been granted and never re-examined.

Binance says roughly 98% of the affected assets had been replenished as of October 31. Those figures come from the company itself, so treat them as reported rather than independently verified. On November 12, the US Justice Department announced the court-authorized seizure of more than $1.2 billion of the stolen funds, with more than $2.7 billion frozen.

What happened at Vanderbilt

On November 9, Vanderbilt University told students and employees to shut down university devices and disconnect from campus Wi-Fi while its IT team responded to an active cybersecurity incident. The university warned that devices, internal systems, and cloud services might be affected. Its help line reportedly went down, and students heard about the incident more than an hour after faculty did.

A ransomware group calling itself BYOD claims it encrypted the university’s databases, took over its domain controller and stole more than one million Social Security numbers. Vanderbilt has not confirmed any of it. The situation illustrates what a real incident looks like from the inside: confusion, unreliable communication, and attackers who understand that uncertainty creates opportunities for phishing.

Vanderbilt is not an isolated case. In five weeks, BYOD has claimed at least nine organizations, including the University of New Mexico Health Sciences Center, where the attack forced the state’s only Level 1 trauma center to divert ambulances, and Michigan State University. Every account of the group’s access points to identity: a credential compromise confirmed at Michigan State, and BYOD’s own claims of a remote access trojan on an employee’s computer and of control over directory services. TrendAI’s companion technical analysis covers how BYOD operates and what defenders can hunt for.

The common thread

The four areas below highlight opportunities to strengthen common security controls:

Credentials and signing authority. A single compromised path to a signing system was enough to move billions. Any high-value action should require approvals that can’t be replayed or pre-staged, and should trigger alerts when they behave unusually.

Developer endpoints. Engineers have broad access and install tools constantly, which makes their machines attractive targets. Software should come from approved, verified sources, not from whatever a search engine returns. Those machines should be monitored like any other privileged system. At Binance, one engineer was compromised through a fake job interview, so recruiting-themed lures belong in the same threat model.

Cloud collaboration. Vanderbilt specifically warned users to watch for unusual permission changes in OneDrive, Box and other cloud storage. Sudden permission changes across shared folders or unusual activity in platforms like OneDrive, Box, or Google Drive can be early warning signs, and they’re often easier to spot than endpoint malware.

Data you don’t need. Bulk personal data is a liability. If an organization doesn’t need to store sensitive data, the safest option is to stop collecting it, or to set strict retention limits.

What to do this month

If you lead security at any organization, a few actions are worth taking now.

  • Map every path that can move money, deploy code, or export sensitive data, and check whether it depends on a single person, device, or approval.
  • Confirm that phishing-resistant MFA protects your privileged accounts, and train people to be suspicious of unexpected login approval requests, which Vanderbilt specifically warned about.
  • Verify that your backups are immutable or offline, and test a restore on a critical system rather than assuming it will work.
  • Make sure your incident communications don’t depend on the systems that might be compromised, because at Vanderbilt, email and the help desk were both affected.
  • Finally, run a tabletop exercise on a ransomware scenario that includes a data-extortion claim you can’t immediately verify, since that’s the situation both organizations were in.

The lesson from these two incidents is not that attackers are unstoppable. It’s that the gap between a trusted path and a compromised one can be very small, and attackers are patient enough to find it.