Purple Teaming: What a Purple Team Does and How to Run One
TL;DR · the short version
Purple teaming is not a third team — it is a way of working where the red team (offence) and the blue team (defence) collaborate in real time instead of operating as adversaries. The red side emulates specific attacker techniques, usually mapped to a framework like MITRE ATT&CK; the blue side watches whether their tooling detects and whether their process responds; and both sides fix the gaps together, there and then. The point is measurable improvement in detection and response, not a win/lose scoreboard. This guide explains what purple teaming is, how it differs from red teaming and penetration testing, when to run one, and a concrete step-by-step method — scope, emulate, observe, tune, re-test, report.
Purple teaming is a security exercise in which the red team (attackers) and the blue team (defenders) work together in real time: red runs a known attacker technique, blue checks whether it was detected, and both fix the gap before moving on. It is a way of working, not a third team, and its goal is a measurable rise in detection and response coverage rather than a win/lose scoreboard.
Planning one? Build the scope, test cards and detection goals with the purple team exercise planner.
Red team vs blue team vs purple team
| Red team | Blue team | Purple team | |
|---|---|---|---|
| Role | Plays the attacker | Defends: monitors, detects, responds | Red and blue working as one |
| Goal | Reach an objective (data, a crown-jewel system) to show you can be beaten | Spot and stop attacks in production | Prove and improve detection for named techniques |
| Style | Adversarial, often stealthy | Continuous operations | Transparent: each technique is announced |
| When findings arrive | In a report at the end | As alerts | Live, technique by technique |
| Output | Attack narrative and findings | Incidents handled | Coverage map plus new or tuned detections, re-tested |
A penetration test is different again: it finds and demonstrates exploitable vulnerabilities in a defined scope, which is about exposure. Purple teaming assumes the attacker is already in and measures whether defenders can see and respond.
How to run a purple-team exercise, step by step
- Scope. Pick one tactic or a small set of techniques, ideally ones used by groups that target your sector. Agree the environment, the time window and the safety rules: what red may and may not touch in production. A threat model helps choose.
- Prepare test cards. For each technique write what red will run, the ATT&CK ID, the log source that should capture it and the alert that should fire.
- Emulate. Red runs each technique in a controlled, announced way, often with open tooling such as Atomic Red Team or MITRE Caldera.
- Observe. Blue checks the stack: did telemetry capture it, did a rule fire, did an alert reach an analyst, and how fast? Record detected, partial or missed.
- Tune. For every miss or partial, add the log source, write or adjust the rule, or fix the threshold, there and then.
- Re-test and report. Red runs the technique again to prove the new detection fires. The report is the coverage list plus the detection content, versioned, with owners for anything not fixed in the room.
Example: mapping an exercise to MITRE ATT&CK
An illustrative credential-access and lateral-movement session on Windows. The technique IDs are from MITRE ATT&CK; the detection columns are what blue checks, not results.
| Technique (ATT&CK ID) | What red runs | What blue should see |
|---|---|---|
| Phishing: Spearphishing Attachment (T1566.001) | A test document with a macro sent to a seeded mailbox | Mail gateway verdict; Office spawning a child process on the endpoint |
| PowerShell (T1059.001) | An encoded PowerShell command | Script block logging (event 4104); EDR alert on encoded commands |
| Scheduled Task (T1053.005) | A scheduled task created for persistence | Windows event 4698 (scheduled task created); EDR persistence alert |
| OS Credential Dumping: LSASS Memory (T1003.001) | Read access to the LSASS process on a test host | Sysmon event 10 (process access) on lsass.exe; EDR credential-theft alert |
| SMB/Windows Admin Shares (T1021.002) | A remote command to a second host over an admin share | Logon events on the target; network detection of admin-share use between workstations |
Each row ends the session as detected, partial or missed, with the rule or log source that changed. Over several exercises that becomes a coverage map you can show to leadership.
Useful next, if this was relevant to you
Metrics that show progress
- Technique coverage: techniques detected out of techniques tested, per tactic.
- Detection quality: detected vs partial (telemetry but no alert) vs missed.
- Time to detect and time to triage for each technique.
- Detections created or tuned in the session, and telemetry gaps found.
- Re-test pass rate: detections from earlier exercises that still fire after tooling and environment changes.
Compare each exercise with your own previous one; that trend is the evidence the programme works.
How often, and who runs it
Treat it as a cadence, not a one-off: a focused exercise on one tactic each quarter is a common pattern, plus re-tests of earlier misses. You do not need a standing red team; an external provider for the day, or an engineer emulating known techniques while the SOC watches, works as long as defenders measure visibility against each technique.
Where it fits with the rest of security operations
Purple teaming feeds detection engineering and incident response directly. Practise writing and tuning rules in the free Detection Lab →, attach a response to each new detection with the IR Playbook Lab, work the resulting alerts in the SOC Case Triage Lab, and use the SOC maturity assessment to see where purple teaming sits on your roadmap. For handing red findings to defenders outside a purple session, see delivering red team reports to blue teams; for the tooling underneath, the SIEM vs SOAR explainer.
Frequently asked questions
What is purple teaming in cybersecurity? A collaborative exercise in which red emulates real attacker techniques, blue checks whether detection and response work, and both fix the gaps together in real time.
What is the difference between a red team, a blue team and a purple team? Red attacks, blue defends; purple is the two working together in the same session, technique by technique.
Is purple teaming the same as a penetration test? No. A pen test is about exploitable exposure in a scope; purple teaming measures and improves detection and response.
Do I need a dedicated red team? No. An external provider for the day, or an engineer emulating known techniques, works.
What framework should it map to? MITRE ATT&CK, so coverage is tracked by technique ID rather than anecdote.
What metrics should it track? Technique coverage, detected vs partial vs missed, time to detect and triage, detections created or tuned, and re-test pass rate.
How often? A recurring cadence, commonly one focused exercise per quarter plus re-tests.
What's the output? Versioned detection content and a measurable rise in technique coverage, not a pass/fail PDF.
How is purple teaming different from red teaming?
A red-team engagement is adversarial and often stealthy: the red team tries to achieve an objective (reach a crown-jewel system, exfiltrate data) without being caught, and reports afterwards. Purple teaming is transparent and collaborative: the red team announces exactly which technique it is running, when, and against what, so the blue team can check their visibility for that specific technique and tune on the spot. Red teaming measures whether you can be beaten; purple teaming measures and improves whether you can see and respond to specific techniques.
Do I need a dedicated red team to run a purple-team exercise?
No. Many organisations run effective purple-team exercises without a standing red team by using an external red-team provider for the day, or by having a security engineer emulate techniques with open frameworks and adversary-emulation tooling while the SOC watches. What matters is that someone executes known techniques in a controlled way and the defenders measure visibility against them — not the size of the team.
What framework should purple teaming map to?
MITRE ATT&CK is the common language: each emulated action maps to a technique (and sub-technique) ID, so coverage can be tracked as "which techniques do we reliably detect?" rather than an anecdote. Threat-informed exercises also use adversary profiles — emulating the techniques a group known to target your sector actually uses — so the exercise reflects realistic risk rather than a generic checklist.
How often should you run purple-team exercises?
Treat it as a recurring cadence, not a one-off. A common pattern is a focused exercise each quarter targeting a specific tactic (for example, credential access one quarter, lateral movement the next), plus a re-test of previously failed techniques to confirm the detections you built still fire after tooling and environment changes. Continuous or automated adversary emulation extends this between manual exercises.
What metrics should a purple-team exercise track?
Track per technique whether it was detected, partially detected or missed, then roll that up into technique coverage (techniques detected out of techniques tested). Add time to detect and time to triage for each technique, the number of detections created or tuned, telemetry gaps found, and the re-test pass rate for detections built in earlier exercises. Compare each exercise with the last, not with an industry figure.
What does a good purple-team output look like?
Not a pass/fail score — a tracked list of techniques with their detection status (detected / partially detected / missed), the specific detection content created or tuned for the missed ones, and a re-test result proving the new detection fires. The deliverable is improved, versioned detection logic and a measurable rise in technique coverage, not a PDF of findings that nobody actions.