Ransomware
Encryption in progress or complete. Containment, scope, and the payment question.
Ransomware is the incident type where the response clock and the investigation clock pull hardest against each other. Every hour spent understanding the intrusion is an hour of encryption or of downtime, and the pressure to restore is enormous and legitimate. The job is to preserve enough to answer the questions that come afterwards, while not slowing the restoration more than the evidence is worth.
The first hour
- Establish whether encryption is still running. That single fact decides whether this is a containment emergency or a forensic one.
- Isolate at the network layer rather than powering off. Powering off destroys memory, which is where the key material, the operator tooling and the injected code live.
- Identify patient zero and the earliest evidence of access. The dwell time before encryption is usually days to weeks, and it is where the actual intrusion is.
- Check whether backups are intact and, critically, whether they are reachable from the compromised network. Backup destruction usually precedes encryption.
- Preserve a ransom note and a sample encrypted file. Both feed family identification and any decryptor that may exist.
Decisions you will be asked to make
Power off, isolate, or leave running?
| Option | What it costs |
|---|---|
| Power off | Stops encryption immediately. Destroys memory, and with it any chance of recovering keys held in RAM. |
| Network isolate | Stops spread and C2 while preserving memory. Encryption already running locally continues. |
| Leave running | Preserves everything and lets the damage continue. Rarely defensible once encryption is confirmed. |
Network isolation is the usual answer because it preserves the highest-value evidence while stopping lateral movement. Power off only when encryption is actively running on a system whose data is worth more than the memory image.
Restore now or investigate first?
| Option | What it costs |
|---|---|
| Restore immediately | Fastest return to service. Destroys the host evidence and often reintroduces the access that caused it. |
| Image, then restore | Adds hours per host. Preserves the ability to answer what was taken and how they got in. |
Image the systems that answer questions and restore the rest. You rarely need every host: patient zero, a domain controller, and one representative encrypted system usually carry the story. Say which you preserved and why.
Was data exfiltrated before encryption?
| Option | What it costs |
|---|---|
| Assume yes | Triggers notification obligations that may not apply. Expensive and hard to walk back. |
| Assume no | Cheap, and wrong often enough to be dangerous. Most modern operators steal before they encrypt. |
| Determine it | Requires egress data you may not have retained. |
Go and look before assuming either way. Firewall, proxy and NetFlow records for the dwell period, plus any cloud storage activity, are the evidence. If the retention window has already rolled, say that plainly rather than reporting no evidence of exfiltration.
Scoping questions
- How many systems are encrypted, and how many are merely reachable from a compromised account?
- When was the earliest confirmed attacker activity, not the earliest encryption?
- Which accounts did they use, and are any of them still valid?
- Were backups accessed, deleted or encrypted, and from which account?
- What left the network during the dwell period?
Collect, beyond the standard triage set
- Memory from at least one still-running affected host, before isolation if it can be done safely.
- A ransom note and one encrypted sample, for family identification.
- Domain controller security logs covering the full dwell period, not just the encryption window.
- Backup system logs and any deletion records.
- Egress telemetry for the dwell period, which is what answers the exfiltration question.
Known mistakes
- Rebuilding patient zero before imaging it. It is the one host that answers how this started, and it is usually the first one someone wants back.
- Treating the encryption timestamp as the incident start. The intrusion is typically days or weeks older, and scoping to the encryption window misses all of it.
- Reporting "no evidence of exfiltration" when what you mean is that egress logs had already rolled. Those are different statements with different consequences.
- Restoring from a backup taken during the dwell period, which restores the access along with the data.