Home/For responders/Working an insider case
Working an insider case
An insider case starts with a signal about someone who is allowed in. The technical work is much like any incident. What changes is who else is involved, what you are permitted to look at, and how quietly the work has to happen.
Before the first query
- Know your authority. The acceptable use policy and the monitoring notice employees agreed to decide what you may review. When a request goes beyond them, or you are unsure, ask Legal before you look.
- Keep the circle small. Name the few people who need to know. Give the case a neutral name that does not include the person's, and keep it out of shared ticket queues and open channels.
- Do not tip anyone off. A changed permission, a disabled account, or a question to a colleague can warn the person. Leave access alone until the case decides otherwise, unless harm is under way, and then contain with HR and Legal agreeing.
- Check for conflicts. The person's manager, close colleagues, and anyone who reports to them should not work the case.
Triage: is there anything here
Most alerts about a person have an ordinary cause: a new project, a changed role, a migration, a script that someone scheduled. Triage answers one question, quietly and quickly: is this out of line with what this person normally does and is allowed to do?
- Compare with their own history. Run the one principal query from the insider data theft template over the previous month, then over the alert window. A person who reads a thousand objects every week is different from one who never has.
- Check whether their access changed. The permission changes query shows grants in the days before. Access granted and then used at once, especially granted by the same person, deserves a closer look.
- Look for the moves that go together. One signal is weak. Notice given, then bulk reads, then data shared outside or copied to a removable drive, is a pattern.
Signals worth combining
| Signal | Where it shows | On this site |
|---|---|---|
| Notice given, or a role ending | HR records only | Agree with HR in advance when they tell security, and what happens next |
| Far more reads than usual | Object-level read logs, which are off by default | Insider data theft, Standard |
| Data shared outside | The cloud's admin audit log | Insider data theft, both levels |
| New ways to decrypt | Key policy and grant changes in the audit log | Insider data theft, both levels |
| Logging turned down | Trail, sink, and exclusion changes | Insider data theft, both levels |
| Files copied to a removable drive | Windows event 4663, with Removable Storage auditing on | Host log shipping, Windows Standard |
| Uploads to personal cloud storage | Web proxy, CASB, or DLP | Not covered here |
| Mail forwarded to a personal address | The mail platform's audit log | Not covered here |
Evidence people can rely on
Insider findings are read by HR, by lawyers, and sometimes by a court. The evidence has to survive someone asking how you know.
- Preserve first. When Legal issues a legal hold, make the systems keep what it names: extend log retention, lock the evidence bucket, and have the mail and document platforms place their own legal holds. A log that expires during a case cannot be recovered, and a legal hold is what prevents it.
- Record custody. For every item: who collected it, when, from where, its hash, and where it is kept. The locked bucket from the evidence template does the keeping.
- Export queries with their context. Save the query text, the time range, and the full results together. A screenshot shows the result without the question that produced it.
- Keep facts apart from conclusions. Write what the logs show: this account read these objects at these times and shared this bucket with that account. Intent is for HR and Legal to weigh.
Handing over
The findings go to the people who decide, in a short memo. It gives a timeline, what data was involved and how much, where it went, what access the person still has, and what has been preserved. If access is to be removed, time it with the HR conversation and remove all of it together: sign-in, sessions, keys and tokens, shared links, and managed devices. Removing some of it early can warn the person before the conversation.
Afterwards
- Look at what they could reach. Access the person had and did not need is the finding the owning teams can act on. The reviewing lesson covers how to ask for it.
- Keep the detections that worked. A query that found this case can become a standing alert, tuned with what the case taught you about normal.
The insider case refresher puts this on one printable page.