Your goal: make a noisy detection program useful again, then turn one detection idea into exact,
tested code, and know what both of them miss.
How this works (about 45 minutes)
- Part one, about twenty minutes, spoken: Marcus asks one detection question and goes deep. He holds
made-up facts about the alerts and the team: ask for what you need.
- Part two, about twenty minutes: a short coding task in the shared editor, Python only. Clarify, say
your approach and its cost, code, run the public tests, then say what breaks it.
- A trade-off at the end, built from your own answers in both parts.
- In part one you can say "can I use the notepad?" to sketch.
- No scores during the call. Your report comes after.
- This is practice built from questions people report for Microsoft. Real rounds vary, so check the
format you are given.
What scores well
- Asking what the noise is made of before fixing anything
- Fixing the noisiest rules at the source, with owners, instead of hiding alerts
- Saying how you'd prove a tuned rule still catches the attack
- In the code: pinning down the key, repeats in the same second and what "steady" means before coding
- Testing jitter at the edge, repeats and mixed-case names before Marcus asks
- A call on the trade-off that weighs the analysts' time against what the rule might catch
Tips
- Take Security Coach · Detection and Security Mock · Brute-Force Detector first.
- Work one tiny log by hand: four connections a minute apart, then the same with one repeat.