Skip to content
All posts

Red Teaming

Red Teaming or Penetration Testing? Choosing the Right Exercise

A penetration test finds what is broken. A red team tests whether anyone notices. How to choose, what sits in between, and what to agree before you sign.

MAMuaaz AhmadPenetration Tester

· 10 min read

On this page

"We want a red team" is one of the most common things we hear at the start of a scoping call. Often the first question back is: what do you want to find out? A penetration test and a red team exercise use many of the same techniques, but they answer different questions. Buy the wrong one and you get a report that answers a question you did not ask.

This post is for whoever holds the testing budget. It explains what each exercise measures, when each fits, the options in between, and what to settle before anyone starts.

Why the two get confused

Both are authorised attacks on your own systems, using real attacker techniques, ending in a report. The difference is what the tester is trying to achieve and how success is measured.

NIST describes penetration testing as security testing in which evaluators mimic real-world attacks to find ways around a system's security features (SP 800-115). The tester's job is coverage. Within the agreed scope, they try to find every weakness that matters, prove it can be exploited and explain how to fix it.

A red team, in the definition NIST's glossary takes from CNSSI 4009, is a group authorised to emulate a potential adversary, with the aim of showing the impact of successful attacks and what works for the defenders (NIST glossary). The operator's job is not coverage. It is to reach a goal, such as "read the finance director's mailbox" or "get to the payments database", while your defenders try to spot and stop them.

A simple way to hold the difference: a penetration test looks at the walls, a red team looks at the guards.

What each exercise actually measures

When I run a penetration test, I want to be noisy if noise helps coverage. I will test every role in the application, try every parameter on the API, and check every host in the range. If the SOC sees me, that is fine. Nobody is grading them. The output is a list of findings, each with evidence, impact and a fix. The API chain we wrote up recently is typical of what a good test should surface: small issues that only become serious once someone joins them up.

When I run a red team, almost the opposite is true. I want to be as quiet as the threat I am emulating. I will pick one route to the objective, not twenty. I will skip vulnerabilities that would get me there faster if using them would look unusual. Every action is logged with a timestamp, so that at the end we can lay my timeline next to the defenders' timeline and see where they matched.

That timeline comparison is the real product. It shows, step by step:

  • which actions raised an alert;
  • which alerts a person looked at;
  • how long it took from first alert to a decision;
  • what the response was, and whether it would have stopped a real attacker.

A red team report will usually have fewer findings than a penetration test of the same environment. That is expected. Its key findings are often about process: an alert that fired but went to a mailbox nobody reads, or a contained laptop with the attacker already on another host.

Side by side

Penetration testRed teamAssumed breachPurple team
Question it answersWhat is exploitable in this scope?Would we detect and stop a real attack?If someone gets in, how far do they get before we react?Does each detection work, and can we tune it?
ScopeDefined systems: apps, APIs, networks, cloudAn objective, plus agreed limitsAn objective from an internal starting pointA list of techniques
Who knowsIT and security teamsA small trusted group onlyA small groupEveryone involved
StealthNot a goalCentralModerateNone, by design
Typical length1 to 3 weeks4 to 8 weeks2 to 4 weeksDays to 2 weeks
Main outputFindings with fixes, then a retestAttack timeline against defender timelineContainment and detection gapsDetections validated or tuned
Best forEvery organisation, before go-live and on a scheduleMature teams with monitoring in placeTeams with monitoring but a smaller budgetTeams building or tuning detection

The durations are typical ranges for a mid-sized environment. Scope drives the real number.

How to decide

Three questions settle most cases.

1. Is anyone watching?

A red team tests detection and response. If you have no SIEM, no endpoint detection, no MDR provider and no one assigned to read alerts, the exercise will reach its objective and the report will say that nothing was seen. You will have paid several weeks of operator time to learn something you already knew.

In that position, a penetration test is the better spend. Fix what it finds and put monitoring in place, whether in house or through a provider. Our post on choosing between an in-house SOC and an MSSP covers that decision. Then test the detection.

2. Have you fixed the obvious?

If your last penetration test found weak service account passwords, unpatched internet-facing systems or an admin panel open to the internet, a red team will probably use one of them. You learn that the same weaknesses still exist, which a retest would have told you more cheaply.

We usually suggest at least one full cycle of penetration testing and remediation before a first red team. The red team should be hunting for what the tests could not show: whether defenders notice and respond.

3. Who is asking, and what will they accept?

The question behind the request often decides the format.

  • A customer, auditor or standard asks for a penetration test. PCI DSS, for example, requires internal and external penetration testing at least every 12 months and after significant changes (requirement 11.4 in PCI DSS v4.0.1). A red team does not replace that, because it does not aim for coverage.
  • A regulator asks for threat-led testing. Some financial regulators run formal red team schemes. In the EU, DORA requires threat-led penetration testing for designated financial entities, and the ECB updated its TIBER-EU framework in 2025 to align with it, including a mandatory purple-team phase. These schemes set the process, who may run the test and the deliverables, so read the rules before choosing a provider.
  • The board wants to know if the company could withstand an attack. This is where a red team or an assumed-breach exercise earns its cost. A finding phrased as "the team reached the objective, and the first alert was handled after the attacker had already moved on" is one a board understands.

The options in between

Many teams do not need the full red team on day one, but have outgrown a test that ignores detection. Two formats sit in the gap.

Assumed breach

We start from a realistic foothold, for example a standard user's laptop and credentials, as if a phishing email had worked. This skips the slowest and least predictable phase, initial access, and spends the budget on what happens next: privilege escalation, movement through Active Directory or cloud identity, and the defenders' reaction.

Skipping initial access is a fair trade: sooner or later someone clicks, and the damage depends on how far an attacker gets afterwards. Phishing and helpdesk testing can be a separate, smaller exercise.

Purple teaming

Here nothing is hidden. We sit with your defenders, run a technique, and check together whether it was logged, whether it alerted, and whether the alert made sense to the analyst. If not, we tune the rule and run the technique again. A day of this often produces more fixed detections than a covert exercise, because the feedback loop is minutes rather than weeks.

We map each technique to MITRE ATT&CK, so the result is a coverage map rather than a list of opinions. Our red team engagements end with this kind of session anyway. Covert operation shows what was missed, and purple teaming fixes it while everyone is in the room.

What good looks like, whichever you choose

Some things should be true of any offensive exercise, and are worth checking in a proposal.

Written rules of engagement. Scope, objectives, systems that must not be touched, testing windows, and who may be contacted. For a red team this includes the "white cell", the small group on your side who know the exercise is running and can stop it.

A safe way to stop. If your SOC escalates to a real incident response, someone must be able to say "this is the exercise" quickly, without exposing the operation to the whole team. Agree how that confirmation works, and how it is verified, before day one. Otherwise an attacker could also claim to be "the red team".

Evidence you can check. A penetration test report gives requests, responses and reproduction steps per finding. A red team report gives a timestamped action log with hosts and accounts used, so your team can find each action in its own logs.

Fixes for both kinds of finding. Technical findings need specific remediation. Detection findings need specifics too: which log source was missing, which rule did not fire, and what would have caught the action.

A retest or replay. For a penetration test, that is a retest of the fixed findings. For a red team, it is replaying the missed techniques once detections have been tuned. Without it you have a report, not a reduction in risk. We covered this in more detail in what a good penetration test delivers.

Questions to ask before you sign

Use these to compare proposals, and to check that you are buying the right exercise.

  • What question will this exercise answer, in one sentence?
  • For a red team: what is the objective, and which threat actor or behaviour are we emulating, and why?
  • What is out of bounds: production changes, physical access, specific people, third-party systems?
  • Who is in the white cell, and how do they stop or confirm the exercise?
  • How will activity be logged so we can match it against our own telemetry?
  • What does the report contain, and can we see a sanitised sample?
  • Is a retest, replay or purple-team session included?
  • For regulated testing: is the provider eligible under the scheme our regulator uses?

When we scope this work, the conversation usually starts with the first question on that list. If the honest answer is "we want to know what is broken", the right exercise is a penetration test, however the request was worded. If the answer is "we want to know whether we would catch someone", and there is a team in place to be tested, that is when a red team is worth the time.

  • Red Teaming
  • Adversary Simulation
  • Pentest
  • Buying Guide
MA

Written by

Muaaz Ahmad

Penetration Tester, FutaqSec

Keep reading

Not sure where to start?

A 30-minute scoping call is enough to understand your environment and recommend the right first step. No obligation.

Book a free consultation