Skip to main content

Automated penetration testing that proves every finding it reports

B-52 automates the whole assessment. It maps your target, writes the test cases that target needs, attempts each one, and reports only the findings it could exploit.

You authorise the scope. The report and the dashboard that come back hold the request, the response and the reproduction steps for each finding, so your engineers start from proof.

Three things are sold as automated penetration testing

The phrase covers a scanner with a schedule, a script that replays a fixed attack sequence, and a platform that plans a test for the application in front of it. They automate different amounts of the work, and the difference shows in four places.

What each kind of automation does at four points in a test
The questionVulnerability scannerScripted automationB-52
What decides the tests A library of known checks, applied to whatever is in scope. A playbook written in advance, run in the same order on every target. A map of this target: its entry points, its roles and its flows. The test cases are generated from that map.
What happens to a suspicious result It is reported as a finding, and your team works out whether it is real. The next scripted step runs if the previous one succeeded. It is a candidate. Exploitation proves it or it is dropped before the report.
What it does with business logic Nothing. A flow taken out of order has no signature to match. Only what the playbook author anticipated. Multi-step flows are run in the wrong sequence and by the wrong role, because that is where logic defects sit.
What reaches your engineers A list sorted by severity. A log of the steps that ran. A report and a dashboard, with the request, the response and the reproduction steps on every finding.

What each kind of automation does at four points in a test

What decides the tests

Vulnerability scanner
A library of known checks, applied to whatever is in scope.
Scripted automation
A playbook written in advance, run in the same order on every target.
B-52
A map of this target: its entry points, its roles and its flows. The test cases are generated from that map.

What happens to a suspicious result

Vulnerability scanner
It is reported as a finding, and your team works out whether it is real.
Scripted automation
The next scripted step runs if the previous one succeeded.
B-52
It is a candidate. Exploitation proves it or it is dropped before the report.

What it does with business logic

Vulnerability scanner
Nothing. A flow taken out of order has no signature to match.
Scripted automation
Only what the playbook author anticipated.
B-52
Multi-step flows are run in the wrong sequence and by the wrong role, because that is where logic defects sit.

What reaches your engineers

Vulnerability scanner
A list sorted by severity.
Scripted automation
A log of the steps that ran.
B-52
A report and a dashboard, with the request, the response and the reproduction steps on every finding.

What B-52 automates, from the first request to the checked report

An assessment runs through six phases in order. Each one is carried out by the platform, and each hands the next something specific.

  1. 1. Discovery

    B-52 reads the target as it stands on the day: hosts, applications, endpoints and the roles that reach them. Client-side code is analysed for routes and parameters the interface never links to.

  2. 2. Planning

    It draws a mindmap of the target and generates the test cases this application needs. Nobody selects checks from a menu.

  3. 3. Scanning

    The generated set is worked through in full. What comes out of this phase is a list of candidates, and none of them is a finding yet.

  4. 4. Exploitation

    Each candidate is attempted inside the authorised boundary. The ones that hold keep the exchange that proved them; the ones that do not are discarded.

  5. 5. Reporting

    The report is written by the platform, and the same findings appear in your dashboard with their evidence attached.

  6. 6. QA

    A quality gate reads the finished report and can send any part of it back to the phase that produced it.

Read how each phase works

Automated inside a scope you sign

A run starts when someone at your organisation signs off the targets and the boundary. In the fully autonomous model that signature is the last thing the run asks of you.

Three kinds of action are held back from the automation until you approve them in writing. They apply in every delivery model.

See the approval gates in full

Changes to production

A destructive or state-changing action against a production system waits for your written approval, even when the host is in scope.

Going past the entry host

Persistence, implants and lateral movement beyond the host B-52 first landed on wait for the same approval. Once you give it, the platform carries the step out itself.

Live credentials and real customer data

Any action that touches either one stops until you have agreed to it in writing.

Proof on every finding

Automation earns its place when the output can be acted on without a second investigation. Each B-52 finding carries three things.

The exchange that proved it

The request that was sent and the response that came back, with the parameter, header or body field that carried the payload. For a mobile, network or directory finding the evidence takes the form that class needs.

Steps your engineers can run

The reproduction steps are attached to the finding in all three delivery models. An engineer can confirm the defect before starting on the fix.

A retest that closes it

When the fix is ready, the same steps are run again against the changed system. A finding closes when the retest can no longer reproduce it.

In the fully autonomous model, findings also pass an independent automated cross-check before they are reported.

The platform has been measured twice. Run in parallel with Security Brigade’s own expert assessment team on the same targets, B-52 reached 90–95% of the combined findings set, which is everything either party found, counted once. B-52 solved all 104 XBOW validation benchmarks, fully autonomously.

How the parallel benchmark was measured

Eleven kinds of target

Web applications, APIs, mobile apps, thick clients, external and internal networks, cloud, Active Directory, social engineering, source code and LLM applications each have a coverage class of their own.

See the eleven coverage classes

Started by a schedule or by a build

A run goes on the frequency you choose, or starts from the pipeline that built the release. GitHub Actions, GitLab CI and Jenkins are in production with customers, and a result can fail a build against a severity threshold you set.

Continuous penetration testing with B-52

Price an automated penetration test for your own target

Pricing for web, API, mobile, cloud, network and source-code assessments is published. Choose the target and its size, and you see the scope and an itemised price before you commit. Other engagements receive a tailored proposal.

Scope and authorisation are confirmed with you before any testing starts.

FAQ

Questions about automated penetration testing

What is automated penetration testing?+

Automated penetration testing is software carrying out the work of a penetration test: finding the attack surface, choosing what to test, attempting exploitation and writing up the result. B-52 automates all six phases of an assessment, from discovery to a QA gate on the finished report.

Is automated penetration testing the same as vulnerability scanning?+

No. A vulnerability scanner matches a target against known checks and reports what matched. An automated penetration test goes on to exploit what it suspects, so every finding it reports has been proved against your system. In B-52, scanning is the third of six phases and its output is a list of candidates for the exploitation phase.

What can B-52 test automatically?+

Eleven coverage classes, from web and mobile applications through to Active Directory, source code and AI applications. Each class has its own page with the methodology it follows.

How long does an automated penetration test take?+

In the fully autonomous delivery model, the observed median from scope sign-off to delivered report is one to three business days.

Can an automated penetration test be used for an audit or a regulator?+

Yes. A B-52 report from any delivery model is signable under Security Brigade’s CERT-In empanelment and usable as audit evidence. The fully autonomous model reports proved findings at an observed median of one to three business days; the expert-verified and human-led models add a senior Security Brigade auditor to the engagement.