Skip to main content

Defending the Persistence Layer

Map nineteen agent controls to the attacks they close, then switch them on and read the logs.

Task 1: Introduction

Task 2: The Execution Layer Controls

Task 3: The Memory Layer Controls

Task 4: The Communication Layer Controls

Task 5: When an Incident Spans Layers

Task 6: Practical Exercise: Before and After

Task 7: Practical Exercise: The Control Panel

Task 8: Conclusion

Back to all walkthroughs
Room Icon

Defending the Persistence Layer

Max room.

Map nineteen agent controls to the attacks they close, then switch them on and read the logs.

medium

50 min

33

User profile photo.
User profile photo.

To access material, start machines and answer questions login.

Across three rooms, Module 4 has named nineteen controls, each closing a different way a compromise outlives the moment it happened: five for code that keeps running (ASI05), seven for memory that keeps being read (ASI06) and seven for messages that keep being trusted (ASI07). None of them were ever switched on. This room puts them side by side against the estate for the first time, and switches on the ones a live incident actually reaches.

The Scenario

This is the TryTrainMe estate you already know: eight agents, with you in the security engineer's seat, as in every room of this module. Nothing new happens here that hasn't already happened somewhere in the three rooms before this one, and you have been reading this estate's logs since the first of them.

An illustrated scene of a security engineer seen from behind at a desk, facing a monitor. Above the desk, a floating network of eight robotic agent heads rests on a flat slab that stands for the shared persistence layer. Every link between two agents carries a green padlock, and a shield emblem sits on the front edge of the slab.

The estate hasn't changed. This room puts a lock on every link.

Learning Objectives

  • Name, for each of the three prior rooms' attacks, the specific control(s) that would have closed it.
  • Distinguish a control that stays on throughout an estate's normal operation from one that only becomes relevant during remediation.
  • Trace a composite incident that spans more than one attack surface, and identify every layer where it could have been stopped.
  • Select the correct subset of controls for a given scenario, rather than reciting an entire catalogue.
  • Classify any control by the moment it acts: write time, trust time, action time, structural or after the fact.
  • Read a guardrail event in an agent log as the reason a step was blocked, and tell an enforced control from one marked NOT ENFORCED.

Prerequisites

This room assumes you've completed all three walkthroughs it integrates. These are hard prerequisites, not background reading: Tasks 2 to 4 reuse their exact controls without re-explaining the attacks behind them.

Answer the questions below

It's time to switch on everything you've already named.