← All work

An Australian financial services licensee · 2026

Security monitoring and evidence retention

Most small firms already pay for Defender and have never switched it on properly. Retention is the one control you cannot apply retrospectively, which is why this is worth doing before you need it rather than after.

The pipeline runs in one line: Defender XDR sends endpoint and identity events to an Event Hub, which passes them to a Function for processing, then to Log Analytics as the queryable store, and finally to Managed Grafana. Two paths branch off that line rather than continuing it. Urgent alerts leave the Function directly for Teams, and the evidence archive is written from Log Analytics to write-once locked storage.

Defender XDR

endpoint and identity

Event Hub

ingest

Function

event processing

Log Analytics

queryable store

32-day detail, 24-month rollup

Managed Grafana

two audiences

Urgent alerts

to Teams

seconds, independent of dashboards

Evidence archive

write-once, locked

7 years, 25 containers

Runs at roughly A$160 a month for the full monitored estate.

The urgent path deliberately does not depend on the dashboards being healthy. If the reporting layer fails, alerting does not.

How to tell whether you have this

Ask two questions. How long are security logs kept, and who looks at them.

If the answer is that your provider handles it, ask them for last month’s. The response to that request tells you what you need to know, and it costs you nothing to make.

Why it happens

Defender licensing comes bundled with the subscription. Nobody is assigned to configure it properly, it is not obviously broken, and nothing forces the question until something forces it very hard indeed.

The asymmetry is what makes this urgent before it looks urgent. The routine question is retention and review. The other one arrives after an incident, from a client or an insurer, and it is some version of when did this start and what else did they touch. If nothing was collecting at the time, that question has no answer and no amount of later effort produces one.

How I would approach it

  1. Establish what your licensing already entitles you to. In most small firms a substantial part of the answer is bought and switched off, so this is configuration and architecture rather than procurement.
  2. Decide retention by purpose rather than uniformly. Investigation and reporting want different things from the same data, and paying to keep everything at investigative fidelity is how monitoring gets declined on cost.
  3. Build the pipeline. Telemetry into a queryable store, with a separate write-once archive for anything that has to survive as evidence.
  4. Split alerting in two. Urgent, event-driven alerts that arrive in seconds and do not depend on the dashboard being healthy, kept apart from a pattern and threshold tier on a daily digest.
  5. Put the tuning behind regression tests. Noise reduction that can silently drop coverage is not noise reduction, and that is the failure mode this work has.
  6. Build for two audiences. An operational view for whoever runs it, and an at-a-glance view for whoever has to answer the question.

What you end up with

Something that can answer a question about last March. That sounds modest until the month somebody asks.

On the engagement below it runs at roughly A$160 a month for the full monitored estate, which is the figure that matters most at this size, because monitoring is usually declined on price rather than on principle.

Seven resilience guards sit underneath it. One of them, an end-to-end delivery canary, paid for itself by detecting a live alerting outage caused by a credential rotation and driving the fix. Without it the first sign of failure would have been a quiet gap in the record found months later.

What it looked like in practice

delivered

Figures from the engagement described above, not a projection of what your estate would produce. Yours will look different.

RefFoundNowWhat the figure refers to
01A$160monthly run cost for the full monitored estate
027 yearsimmutable evidence retention, across 25 write-once containers
034 + 3adversarial reviews and compliance passes cleared before build started

Is this yours?

If any of the above sounds like your estate, say so and I will tell you what checking would involve. Most of these start as a short look rather than a project.

Not sure it is this one? The rest of the work covers monitoring, identity, deployment, cost, AI use and compliance documentation.

Related work

All work