A purple team exercise without a written plan tends to produce a pleasant day of "we saw some of it" and very little that changes detection coverage. The value comes from agreeing in advance which adversary behaviours will be emulated, what telemetry and alerts each one should produce, what counts as a pass, how the exercise will be stopped safely, and how gaps become tracked engineering work. This template gives [Organisation] a plan structure for exactly that, with test cases mapped to MITRE ATT&CK. It is for detection engineers, SOC leads, red and purple team operators and the risk owners who approve testing in production. The concepts and operating model are covered in the Purple Teaming Guide; this page is the document you fill in.
How to use this template
- Replace text in square brackets, such as [Organisation], [Exercise lead] and [SOC], with your own names, systems and dates.
- Keep sections 1, 2 and 4 for approval by [Risk owner] before any test runs. Sections 3 and 8 are working records completed during and after the exercise.
- Use only emulation content from reviewed sources, such as Atomic Red Team tests or MITRE Center for Threat-Informed Defense emulation plans, run by authorised operators on in-scope systems. This template deliberately contains no attack commands.
- Prefer several short, focused exercises (for example one ATT&CK tactic or one threat actor's key behaviours per session) over a single long event.
- Store the completed plan and results with restricted access: they describe exactly where detection is weak.
1. Scope and triggers
1.1 Purpose and objectives
This exercise validates that [Organisation]'s detection and response controls identify and respond to the adversary behaviours listed in section 3. Objectives:
- Confirm that the telemetry required for each test case is collected and searchable in [SIEM].
- Confirm that detection rules fire with the expected severity and reach the right queue.
- Measure time to detect and time to triage for each test case.
- Produce a prioritised list of detection, logging and control gaps with owners.
1.2 Triggers for this exercise
[Select those that apply: scheduled quarterly cycle; new threat intelligence on [threat actor] targeting our sector; new EDR, SIEM or logging platform; findings from a recent incident; new detection rules needing validation; audit or regulatory requirement.]
1.3 Threat basis
Behaviours are selected from [threat actor or scenario, for example a ransomware affiliate playbook or a cloud credential theft scenario], based on [intelligence source]. Where a full emulation plan exists in the MITRE Center for Threat-Informed Defense Adversary Emulation Library, reference it; that library provides full plans for named actors and micro emulation plans for compound behaviours such as web shells.
1.4 In scope and out of scope
- In scope: [named test hosts, test user and service accounts, cloud test account or subscription, network segments].
- Out of scope: [production databases, safety or OT systems, customer data, third-party systems not covered by written authorisation].
- Prohibited actions: real data exfiltration, destructive actions, encryption of real data, disabling production security tooling, denial of service, and use of real customer or employee credentials.
- Window: [dates and hours, time zone], with a change record [reference].
2. Roles and contacts
| Role | Responsibility | Name and contact |
|---|---|---|
| Exercise sponsor and risk owner | Approves scope and accepts residual risk | [Name, phone] |
| Exercise lead | Runs the schedule, keeps the log, calls pauses | [Name, phone] |
| Emulation operators (red) | Execute approved test cases only, record exact times | [Names] |
| Detection engineers and SOC analysts (blue) | Observe telemetry and alerts, triage, tune rules | [Names] |
| System owners | Confirm test hosts and accounts, watch for business impact | [Names] |
| MSSP or MDR contact | Aware of the exercise; follows agreed alert handling | [Provider, contact] |
| Abort authority | Can stop the exercise at any time | [Exercise lead and on-call security manager] |
- White cell decision: state whether the SOC on shift is told in advance (open book, the usual purple format) or not (to test real response). If not told, a named white cell member must be able to confirm "this is the exercise" within minutes to prevent a real incident response.
- Deconfliction: [Exercise lead] checks with [SOC lead] before each test case that no real incident is in progress on the same systems.
3. Detection and triage
This section is the core of the plan. Each test case states the ATT&CK technique, the emulation source, the telemetry that should exist, the detection expected and the pass criteria. Agree the expected results before execution, not after.
3.1 Test case register
| ID | ATT&CK technique | Emulation source | Expected telemetry | Expected detection | Pass criteria |
|---|---|---|---|---|---|
| TC-01 | T1059.001 PowerShell | [Atomic test name and GUID] | PowerShell script block logging (Event ID 4104); process creation (4688 or Sysmon Event ID 1) | [Rule name] for suspicious encoded or obfuscated PowerShell | Telemetry present within [5] minutes; alert raised at [High]; triaged within [30] minutes |
| TC-02 | T1003.001 LSASS Memory | [Atomic test name and GUID] | Sysmon Event ID 10 process access to lsass.exe; EDR credential theft telemetry | EDR alert and [SIEM rule] | Blocked or alerted; analyst identifies the source process and account |
| TC-03 | T1053.005 Scheduled Task | [Atomic test] | Event ID 4698 scheduled task created; process creation | [Rule name] for tasks created by non-admin tooling | Alert raised; persistence location recorded in triage notes |
| TC-04 | T1547.001 Registry Run Keys | [Atomic test] | Sysmon Event ID 13 registry value set | [Rule name] | Alert or hunt query returns the change within [15] minutes |
| TC-05 | T1087.002 Domain Account discovery | [Atomic test] | Process creation; directory query logs where available | [Rule or hunt query] | Activity visible in hunt query; alert optional (document decision) |
| TC-06 | T1021.001 Remote Desktop Protocol | [Test procedure between two test hosts] | Event ID 4624 logon type 10 on target; network flow logs | [Rule for RDP from workstation to workstation] | Alert raised with source and destination hosts |
| TC-07 | T1098.001 Additional Cloud Credentials | [Cloud test in test account] | Cloud audit trail event for new access key or service principal secret | [Cloud detection rule] | Alert within [10] minutes naming the identity and the new credential |
| TC-08 | T1567.002 Exfiltration to Cloud Storage | [Test using dummy data only] | Proxy or DLP logs; EDR network telemetry | [DLP or proxy rule] | Upload of canary file detected; no real data leaves |
| TC-[n] | [Technique ID and name] | [Source] | [Data source and event] | [Rule] | [Measurable criteria] |
3.2 Result recording
For each test case record: execution start and end time (UTC), operator, host and account used, whether telemetry was present (yes, partial, no), whether the alert fired, alert time, triage time, analyst notes, and the result.
| Result | Definition |
|---|---|
| Detected | Alert fired as expected and was triaged within the target time |
| Detected late | Alert fired but outside the target time, or at the wrong severity or queue |
| Telemetry only | Data was collected and found by a hunt query, but no alert fired |
| Prevented | A control blocked the action; record whether an alert was also raised |
| Not detected | No alert and no usable telemetry |
4. Containment
In this plan, containment covers two things: validating response actions, and keeping the exercise itself under control.
4.1 Response actions to validate
- [SOC] isolates the test host through EDR within [15] minutes of a High alert, and the isolation is confirmed.
- [Identity team] disables the test account and revokes its sessions and tokens; the result is verified by a failed sign-in.
- [Cloud team] deactivates the test access key created in TC-07 and confirms no further API calls succeed.
4.2 Safety controls and stop conditions
- The exercise pauses immediately if any of these occur: unexpected impact on a production service, activity observed outside the agreed scope, a real security incident is declared, or a test produces results the operators do not understand.
- Any participant may call "stop" through [channel]; the exercise lead confirms the pause and logs the time.
- Operators record every action with timestamps so that blue team analysts can match activity to alerts and any unexpected effect can be traced.
- Tests flagged as requiring elevation (Atomic Red Team tests carry an
elevation_requiredfield) are run only on dedicated test hosts.
5. Eradication
Every test leaves artefacts that a real attacker could later reuse or that could hide real activity. Clean up before the window closes.
- Run each test's documented clean-up step (Atomic Red Team tests define a
cleanup_command) and confirm the artefact is gone. - Remove test accounts, access keys, scheduled tasks, registry values, files and firewall rules created during the exercise, using the operator log as the checklist.
- Rotate or delete any credentials used by operators during the exercise.
- Confirm with [SOC] which alerts were exercise-generated and close them with an exercise tag, so they do not distort metrics or hide later real alerts.
6. Recovery
- Return test hosts to their baseline image or snapshot, and confirm EDR and logging agents are healthy.
- Revert any detection rule changes made live during the exercise that were not approved through change control, or raise the change formally.
- Release isolated hosts and re-enable test accounts only if they are needed for retesting.
- Confirm with system owners that no residual impact remains, and close the change record.
7. Communication and notification
- Before: notify [SOC lead], [MSSP or MDR], system owners and the IT service desk of the window, the source hosts and accounts, and the white cell contact. Agree a code phrase for confirming exercise activity.
- During: use [dedicated channel] for real-time coordination; the exercise lead posts the start and end of each test case.
- Escalation: any suspicion of real malicious activity is escalated through the normal incident process immediately, and the exercise pauses.
- After: circulate a short summary within [two] working days and the full report within [two] weeks to [Security steering committee], restricted to need-to-know.
8. Post-incident review
8.1 Exercise review
- Hold a joint red and blue review within [five] working days, walking through each test case result.
- For every result other than "Detected", record the root cause: missing log source, parsing failure, rule logic gap, alert routing, analyst playbook, or control misconfiguration.
- Raise an engineering ticket per gap with owner, priority and target date, and link it to the test case ID.
8.2 Metrics
- Percentage of test cases Detected, Detected late, Telemetry only, Prevented and Not detected.
- Median time to detect and time to triage across test cases.
- Number of gaps closed and verified by re-test since the previous exercise.
- ATT&CK technique coverage change, mapped in ATT&CK Navigator or your coverage tool.
8.3 Re-test
Re-run every failed test case after its fix, and add passing test cases to an automated regression suite so detections are checked continuously. See the Breach and Attack Simulation (BAS) Guide.
Implementation guidance
- Start small: a first exercise of six to ten test cases on two test hosts and one cloud test account is enough to expose logging and routing gaps.
- Choose techniques from threat intelligence: prioritise behaviours used by actors that target your sector, and techniques where you have no tested detection. See the Threat-Informed Defence Guide.
- Use reviewed emulation content: Atomic Red Team provides a library of small tests mapped to ATT&CK under the MIT licence, executable directly or through the Invoke-Atomic framework; Caldera automates adversary emulation on ATT&CK; the Center for Threat-Informed Defense publishes full and micro emulation plans. Review every test before use, because some download tools or need elevation.
- Write expected results first: filling the telemetry and pass criteria columns before execution is what turns the session into a measurable test.
- Get written authorisation: section 1 and 2 signed by the risk owner, aligned with your Red Team Rules of Engagement Template.
- Include identities: add test cases for service account misuse, token theft and new cloud credentials, not just endpoint techniques. See the Detection Engineering for Non-Human Identities Guide.
- Repeat on a cadence: quarterly for core techniques, plus ad hoc runs after major platform changes.
| Framework | Reference | How this plan supports it |
|---|---|---|
| NIST CSF 2.0 | ID.IM-02, DE.AE-02 | Improvements identified from security tests and exercises; analysis of adverse events |
| NIST SP 800-53 Rev. 5 | CA-2, CA-8, IR-3, RA-10 | Control assessments; penetration testing; incident response testing; threat hunting |
| MITRE ATT&CK v20 | Technique IDs in section 3 | Common language for test cases and coverage |
Standards and references
- MITRE ATT&CK
- Red Canary: Atomic Red Team
- MITRE Center for Threat-Informed Defense: Adversary Emulation Library
- MITRE: Caldera
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: Purple Teaming Guide · Detection Engineering Guide · Red Team Rules of Engagement Template · Tabletop Exercise Guide