This class is an assessment, not a training exercise. It is built from reconnaissance of your own organisation rather than from a template library, it runs against a population you authorised in writing, and it is judged on one thing: whether it obtained a way in. What comes back is the route, the evidence, and what your own controls did about it.
The boundary
Every coverage class has three actions that wait for written approval. This one has something before them: a document that must exist before anything is sent to anybody.
| State | What it means | What follows |
|---|---|---|
| Sending anything to a real person | Nothing reaches anybody until a written authorisation naming the population, the window and the permitted pretexts exists. It cannot be released mid-run, because it is what makes the engagement an engagement. | Required before the work begins. |
| Using what a captured credential opens | A credential that was entered is the evidence. Signing in with it is the next act, and it is the third approval gate under a different name. | Stops and waits, in writing. |
| Persistence and movement past the entry host | Where a pretext yields access to something, going further into the estate is an internal network question rather than a social engineering one. | Stops and waits, in writing. |
| Everything else inside the authorised scope Terminal | Open-source reconnaissance, target enumeration, campaign construction, delivery inside the agreed window, and reporting. | Runs without asking. |
What the authorisation names
The population, by group rather than by whoever reconnaissance happens to surface. The window, so your own team can tell this apart from a real campaign afterwards — and so a genuine incident during it is not filed as ours. The pretexts that are permitted and the ones that are not, which is your decision and one most organisations turn out to have firm views about the moment they are asked. And where the engagement stops: at the credential being entered, or at what the credential opens. In the fully autonomous model that document is the last human action of the engagement, which is precisely why it has to be this specific.
Class and boundary
Reconnaissance is not preparation for the other two. It is the form that most often produces the finding.
Who is there, what is already published about them, which of their credentials are in circulation somewhere nobody put them deliberately, and which of that makes a campaign credible. It runs first and it decides the other two.
A campaign written for the organisation it is aimed at, out of what reconnaissance actually found — not a template with a different logo dropped into it.
A scenario with a plausible reason to exist, run to obtain a credential or an action, and treated as one link in a chain rather than as the result.
Where a pretext or a recovered credential yields access, how far that access then travels is the internal network class. This one ends at the way in; that one measures what the way in was worth.
Telephone pretexting sits outside this class. Physical, hardware and wireless testing are out of scope for the platform entirely, in every class.
The other thing this could be
A security awareness programme runs phishing simulation to measure and improve how a workforce responds: at a cadence, repeated, with everybody eventually included. That is a real discipline and Security Brigade runs one. This class is doing a different job — it is trying to obtain a foothold, against a named population, once, as part of an assessment, and it succeeds or fails on whether it got in rather than on what proportion of people clicked. The two produce different reports for different readers. Buying either one does not give you the other, and the failure mode of confusing them is a programme that never tests what an attacker would actually do, and an assessment that quietly turns into a staff appraisal.
Who this page is for
Both are answered here. Where they should start is not the same.
The run
The phase names are the platform’s. On this class the first one is where most of the value is, and the fourth is the only one anybody outside the engagement ever sees.
Phase one
Reconnaissance: who is in the authorised population, what is published about them and by them, and which of their credentials are already circulating in corpora assembled somewhere else entirely. Findings from this phase cost the campaign nothing and sometimes make it unnecessary.
Phase two
The authorisation: population, window, permitted pretexts, and where the engagement stops. Signed before anything is written. In the fully autonomous model it is the last human action of the engagement.
Phase three
Campaign construction and rehearsal — the pretext, the infrastructure it comes from, and whether your own mail controls stop it. A campaign that never arrives is a result about those controls, and it is reported as one.
Phase four
Delivery inside the agreed window. What comes back is a credential entered, an action taken, or nothing at all — and nothing is a result too, with the reasoning behind it written down rather than left as silence.
Phase five
Each interaction recorded with its timestamp and the pretext behind it, landing as it happens rather than at the end, so your own team can correlate it against what they saw at the time.
Phase six
The last phase runs on the finished report and can return it. It is where an interaction is separated from an outcome — somebody opening a message is not the same fact as somebody giving something up.
On the people in it
An engagement like this generates a list of people who did something, and what happens to that list decides whether running it was worth anything. The report describes the population, the pretext and the outcome; it does not hand a manager a league table of names. Where you need individual detail for a specific reason, that is agreed in the authorisation with everything else and it is your decision rather than a default. The findings worth having are structural in any case: a pretext that worked says something about what your staff have been trained to expect, and a captured credential that opened something says rather more about what a credential is permitted to open.
What the assessment returns
The first two rows are the ones that survive into a remediation plan. The rest are what explain them.
| Class | What it looks like in a real engagement |
|---|---|
| Credentials already in circulation | Accounts belonging to the organisation that appear in corpora assembled elsewhere — from unrelated breaches, from software running on somebody’s personal machine, from a repository that was made public by accident. Found before anything is sent, and they cost the campaign nothing. |
| What a credential opens | Whether a recovered or captured credential reaches anything at all, and what stands between it and the account. This is the finding. The credential is only the evidence for it. |
| The published surface of the organisation’s people | Who is identifiable, in which role, with which address format, and what has been published about the processes they take part in. The material any campaign is built from, and all of it already public. |
| What the pretext relied on | Which assumption made the scenario credible — an expected process, a known supplier, a system people already receive mail from. This is the part that turns into a change somebody can actually make. |
| What the mail controls did | Whether the campaign arrived, was quarantined, or was stopped outright. Reported either way, because a campaign that does not arrive is evidence about a control you paid for. |
| What happened after the interaction | Whether the destination was reachable, whether anything warned the person, and whether what they did could be undone. |
| Whether anybody said something | Whether the activity was reported internally at all. That is the half of the exercise that describes the organisation rather than the campaign. |
Credentials already in circulation
What a credential opens
The published surface of the organisation’s people
What the pretext relied on
What the mail controls did
What happened after the interaction
Whether anybody said something
Three of them, worked
Each of these reads one way in a summary and another way in a remediation plan.
Methodology
The last row is applied to the technical findings a campaign produces rather than to the campaign itself, and the row says so rather than pretending a scoring vector fits everything.
| Standard | Version | What it carries here |
|---|---|---|
| MITRE ATT&CK | v19.2, released 6 August 2026. Read 2026-09-13 | The technique vocabulary for reconnaissance, initial access and credential access — so a finding can be handed straight to the team that owns the detection for it. |
| NIST SP 800-115 | Final, September 2008. Read 2026-09-13 | The structure of a technical assessment, and the handling rules for what the attack phase turns up. |
| Still the current edition: it has been neither withdrawn nor superseded. | ||
| PTES | Current | The engagement structure — intelligence gathering, authorisation in writing, and the shape of the report. |
| CVSS | v4.0, November 2023. Read 2026-09-13 | The severity vector on the technical findings a campaign produces — what a captured credential reached, and what did not stand in its way. |
| Not applied to the campaign itself. A pretext is not a vulnerability with a vector, and scoring one as though it were would be inventing precision. | ||
MITRE ATT&CK
NIST SP 800-115
Still the current edition: it has been neither withdrawn nor superseded.
PTES
CVSS
Not applied to the campaign itself. A pretext is not a vulnerability with a vector, and scoring one as though it were would be inventing precision.
Before a run starts
Deliberately excluded
Per finding
The message as delivered and the infrastructure it came from, so your team can find it in their own logs rather than taking our word for what arrived.
The interaction and its timestamp, described by group and role. Individual detail appears only where the authorisation asked for it.
What a captured credential authenticated to, and what stood in the way. This is the finding the engagement exists to produce, and it is also the one behind an approval gate.
Whether the campaign arrived at all, what happened to the parts that did not, and whether anybody inside the organisation raised it.
In the model with no auditor in it, findings pass a second automated gate before they are reported. It is a check on top of the evidence, not a substitute for it.
Where this sits
No vendor is named here — these are categories of work, and one of them is a thing Security Brigade also sells.
| An awareness programme | Mail gateway and filtering | A scheduled red team | B-52 | |
|---|---|---|---|---|
| What counts as success | A measurable change in how people respond, over time. | A campaign that never arrives. | A foothold. | A foothold, or a written account of what stopped there being one. |
| Who it is aimed at | Everybody, repeatedly, by design. | Everybody, continuously. | A chosen few. | The population your authorisation names, once. |
| Built from reconnaissance of your organisation | Usually from a template library, which is appropriate for the job it does. | Not applicable. | Yes. | Yes — and the reconnaissance is reported whether or not the campaign ever ran. |
| Continues past the credential | No, and it should not. | Not applicable. | Yes, within scope. | Only where you authorised it, and the report says which findings went that far. |
| Cadence | Continuous, by design. | Continuous. | When it is scheduled and staffed. | The interval you set. |
| Whose signature it carries | Not that kind of output. | None. | The firm that ran it. | Security Brigade’s, in the two models with an empanelled auditor in them. |
What counts as success
Who it is aimed at
Built from reconnaissance of your organisation
Continues past the credential
Cadence
Whose signature it carries
On the awareness programme
Running phishing simulation as a training programme should include everybody, should repeat, and should be measured over months. None of that is true of this class, which aims at an authorised population once and is judged on whether it got in. Both are worth having and Security Brigade runs both, on separate terms with separate reports. What neither of them should be is the other: a training programme asked to prove whether an attacker could get in will not, and an assessment asked to produce a departmental score has stopped being an assessment.
The record
Coverage is identical across the three delivery models. What differs is whose signature the report carries.
A social engineering engagement is testing performed on people. The written authorisation naming the population, the window and the permitted pretexts is what makes it an engagement, and it is retained as part of the record rather than treated as paperwork.
Where an obligation asks for testing that covers the route into an organisation through its people, this class produces it — the campaign, the outcomes and the authorisation that permitted them, together.
A finding here closes on a control change rather than on a patch: a second factor that now stands in the way, a process that can now be verified, a credential that no longer works.
Where the report goes to a regulator or an assessor, start with the expert-verified model.
Worked example
Four steps, and the first of them was found in reconnaissance before anything was sent to anybody.
| Link | Alone | In sequence |
|---|---|---|
| 1 A credential already in circulation | An address and a password, in a corpus assembled somewhere else. | It belongs to somebody inside the authorised population. |
| 2 An account asking for one thing | A sign-in page. | The credential works, and nothing further is asked for. |
| 3 A mailbox the organisation trusts | An inbox. | A message sent from it does not look like a campaign, because in every technical sense it is not one. |
| 4 The action somebody took | A routine request, actioned. | The route in never touched the perimeter, so nothing watching the perimeter saw it. |
Measured
B-52 was run beside Security Brigade’s expert assessment team on the same targets at the same time, and everything either of them produced was pooled into a single set with each item counted once. B-52 reached 90–95% of that pooled set. Part of what it reached the team had not, which is why the pool is larger than either side alone — and why the expert-verified model is a genuine option on this class in particular, where judgement about what to send is worth having a person on.
One scan is one application or target and the entry tier is $500, but this class is scoped around a population, a window and a written authorisation. That is a conversation, and it is a short one.