2.3.4. Attack Simulation Training
💡 First Principle: You can't measure how phishable your users are by asking them. A realistic, harmless simulation measures it, and immediate training at the moment of failure is when the lesson sticks.
Attack simulation training (Defender portal > Email & collaboration > Attack simulation training) requires Defender for Office 365 Plan 2 or Microsoft 365 E5. It's run by Global Administrators, Security Administrators or the dedicated Attack Simulation Administrator role; an Attack Payload Author can create payloads but can't launch simulations.
Simulation techniques (seven):
| Technique | What the simulated user is lured to do |
|---|---|
| Credential Harvest | Enter credentials on a fake sign-in page |
| Malware Attachment | Open a (harmless) attachment |
| Link in Attachment | Click a link inside an attachment |
| Link to Malware | Click a link that downloads a file |
| Drive-by URL | Visit a page that would run code |
| OAuth Consent Grant | Grant permissions to a malicious app |
| How-to Guide | Training material that teaches users to report phishing |
The building blocks:
- A payload is the link, QR code or attachment in the simulated message. Payloads can be built-in, custom, or harvested from real phish your users reported (payload automations).
- Login pages are used by Credential Harvest and Link in Attachment simulations.
- Simulation automations reuse a configuration for repeated runs on a schedule.
- Training campaigns assign training modules without a phishing simulation — for example, annual awareness training for all staff.
- The predicted compromise rate estimates how many users are likely to fall for a payload, so you can pick a realistic difficulty.
⚠️ Exam Trap: If a requirement is "assign security awareness training to all employees without sending a phishing test," the answer is a training campaign, not a simulation. And if the tenant has only E3, attack simulation isn't available at all — it needs Plan 2.
Reflection Question: After a real credential-phishing incident, you want to test whether users would fall for the same lure and train those who do. Which feature lets you reuse the real attacker's message, and which technique would you choose?