On Monday morning, Vanderbilt University's IT team sent out an instruction most incident responders treat as a last resort: shut down your device, disconnect from Wi-Fi, do not try to fix this yourself. Students watched desktops fill with copies of a file called README_BYOD_31.txt. Some faculty saw their documents rewritten with a .byod extension. BleepingComputer obtained a photograph of one such screen and confirmed the contact link in the note resolves to a leak site run by a group calling itself BYOD.
The group's claim is expansive. From its leak page: "We have your entire estate and research data. Over 1 Million SSNs, Full Names, Email addresses, Phone Numbers, Student & Employee Records, Full takeover of your DC, whatever was on your network, we possess." Vanderbilt has confirmed an active cybersecurity incident. It has not confirmed encryption, data theft, or that BYOD is responsible. Those are three separate questions on three different timelines, and conflating them is how response plans fall apart.
What is actually corroborated
The .byod extension is the useful fingerprint. Nashville Scene reported the same extension showed up at a New Mexico hospital and at Michigan State University in the two weeks before Vanderbilt. That is a pattern, not a coincidence.
Victim counts are less reliable. BleepingComputer says BYOD launched in 2026 and currently lists five other organizations besides Vanderbilt. The Scene, citing the Ransomware.live tracker, says the group has hit 17 targets since September. Leak sites prune and inflate entries routinely, so treat any number on a ransom page as marketing until a victim confirms it.
Vanderbilt's own advisory listed the signs to watch for: unexpected pop-ups, unusual permission changes in cloud storage, missing or modified files, unfamiliar applications, and unexpected authentication requests. Most of that is boilerplate. The authentication requests are the one worth circling. If the domain controller claim has any truth to it, the attackers already have credential material and the interesting activity is happening in your identity provider, not on disk.
Why a domain controller changes the response
If "full takeover of your DC" is accurate, the mass shutdown order makes sense. A domain controller means Group Policy, which means one edit pushes a logon script or scheduled task to every machine in the domain. It means files can land on thousands of hosts without anyone touching them individually. Encrypting a file server does not produce laptops spontaneously filling with ransom notes. A GPO, a remote execution sweep, or a management agent does.
Cloud sync adds a second delivery path. Encrypt a file locally, let OneDrive or Dropbox upload it, and the encrypted version overwrites the good copy in the cloud. Then the next machine syncs down. This is exactly why Vanderbilt told people to close Word documents and OneDrive and Dropbox folders rather than just unplug the Ethernet cable. For a university with research data spread across personal cloud accounts and shared drives, the sync client is the propagation vector.
The cost of that advice
Shutting down a machine destroys volatile memory. Running processes, network connections, injected code, and sometimes the encryption key itself live there and nowhere else. Hibernation files go too. If you kill power, you lose the process tree that would have told you which binary did the encrypting and how it got there.
The better instruction, when time allows, is to disconnect from the network and leave the machine running, then call the service desk. Vanderbilt had no time. When you cannot tell whether persistence lives in a GPO on a domain controller or a scheduled task on four thousand laptops, a mass shutdown buys containment and you pay for it in forensic artifacts and recovery speed. I would have given the same order. That does not make it free.
The comms problem nobody budgets for
The Hustler reported that some professors canceled classes while the university's official statement said classes were proceeding as scheduled. That gap is small and it is expensive. Faculty who received no guidance and improvised will trust their own judgment the next time an IT email arrives, including the phishing one that inevitably follows a breach. Write the message before you need it. Vanderbilt's line, "DO NOT attempt to investigate, troubleshoot, or resolve the issue yourself," is a good template because it removes the option of a well-meaning graduate student making things worse.
Also note the negotiation channel: SESSION, not email. That means evidence collection and negotiation both depend on a messenger platform the university's legal and IT teams have probably never used. Rehearse that part too.
What to check this week
- Can your identity provider revoke every active session at once? Disabling accounts is not the same thing. Kill the tokens.
- Does your EDR know how to quarantine a syncing folder without deleting the local copy? Test it on one machine.
- Are your domain admin accounts logging into endpoints that browse the web? A DA session on a laptop with an infostealer is a DA session an attacker owns.
- Do you have a backup copy that no domain account can reach, stored in a separate provider account with its own credentials? A DC compromise should not hand over your recovery path.
- Have you run a tabletop that treats leak-site extortion separately from encryption? The clocks and the decision makers are different.
BYOD also took credit for an alleged attack on the Texas Health and Human Services Commission the same day, according to DataBreach.com. Two universities and a hospital inside three weeks, all with the same extension, is a campaign by any reasonable definition. The question for every other IT department is whether their domain admin credentials are sitting on an endpoint that reads email.