Skip to content

Penetration Testing

Know which flaws put your business at risk.

A source-assisted test of your web applications, APIs and cloud perimeter. We prove each finding from outside, rank it by business impact and retest fixes as agreed in scope.

What we test

Web applications, APIs and cloud perimeter, from an attacker’s position with your source code in hand.

What you get

Confidential report and letter of attestation, plus retest records for retesting agreed in scope.

How to start
  1. 1Scoping call with a practitioner
  2. 2Written scope and approvals
  3. 3Testing in an agreed window

Coverage references: OWASP WSTG 4.2 · OWASP ASVS

What we test

Web applications, APIs and cloud perimeter, from an attacker’s position with your source code in hand.

What you get

Confidential report and letter of attestation, plus retest records for retesting agreed in scope.

How to start

Scoping call → written scope and approvals → testing in an agreed window

Scope

Every boundary an external attacker could cross.

We find where sign-in, permission and tenant checks live in your code, then test the routes, parsers and background jobs that could get around them.

Web applications

  • Sign-in, recovery, invitations, SSO and MFA
  • Sessions: rotation, expiry and revocation
  • Permissions and privilege escalation
  • Isolation between customer tenants
  • Business logic: approvals, entitlements, payments
  • XSS, CSRF, CORS and framing
  • File upload, download and processing

APIs

  • Identity and tenant binding on every REST, GraphQL and RPC route
  • Object, field and function-level access
  • Mass assignment and over-exposed fields
  • Bulk, export and search operations
  • Webhooks, URL imports and server-side fetches
  • Injection through queries, templates and paths

Cloud perimeter

  • Exposed origins, admin and debug interfaces
  • Ingress, public identities and service accounts
  • Public storage and signed URLs
  • Transport, caching and error configuration
  • Deployed dependencies that are actually reachable
  • Data flows to third parties

Not included: internal corporate networks and endpoints, phishing and social engineering, physical testing, continuous monitoring.

Process

Nothing is tested until you approve it in writing.

AI agents run much of the code review and testing, inside the limits you approve. A named tester of record is responsible for the report and signs the letter of attestation.

StepHexkeyYour team
1 · Scope

Agree goals, applications and environments.

You: Name an authorizing person and a clarification contact.

Approval 1 · reviewWhich repositories, branches and cloud accounts we may read.
2 · Investigate

Map architecture and attack paths from source and cloud metadata. No requests reach your application.

You: Answer questions as they come up.

3 · Plan

A test plan for your application: tests, accounts, limits and the evidence each needs.

You: Review the plan.

Approval 2 · testingTargets, testing window, request limits, test accounts and an emergency contact.
4 · Test

Bounded tests from an attacker's position. Serious issues can be raised before the final report.

You: Keep the emergency contact reachable.

5 · Report

Confidential report and letter of attestation, signed by your tester of record.

You: Fix, then request a retest as agreed in scope.

What we need from you

  • Read access to named repositories, or a source export
  • Cloud metadata access, or a decision to leave it out
  • The environment to test and its deployed release
  • Two test tenants, with accounts for each role in scope
  • Safe test data, and an owner for clean-up

How we test production

  1. BEFOREWe agree request, rate and concurrency limits, and the test accounts and data we’ll use.
  2. DURINGYou sign in to the test accounts yourself, so we never handle passwords, MFA codes or tokens. We pause if you ask, or on instability, unexpected customer data or lockouts.
  3. AFTERCredentials are redacted before evidence is stored. Anything outside scope is recorded and left alone.

Deliverables

Findings an executive can act on and an engineer can reproduce.

Each finding gives leadership the impact and the urgency, and gives engineering the root cause, a redacted reproduction and the fix.

Sample finding · from a confidential reportFictional · not client data
HXK-0000-FND-0001 · CWE-639

Any customer can read another customer’s user records

Severity: CriticalPriority: Fix first
For leadership
What an attacker gains

Any signed-in user of any customer account can read the names, email addresses and roles of every other customer’s users.

Who is affected

All customer tenants.

Why fix first

It needs only an ordinary account and one changed number in a request.

How it was validated

Reproduced from outside with two controlled test tenants. No real customer data was accessed.

For engineering
Root cause

Tenant scope is checked on the list route but not on GET /api/v1/users/:id.

# Signed in as test user A (tenant A)
GET /api/v1/users/7731 HTTP/1.1   # user in tenant B
Host: api.example.test
Authorization: Bearer [REDACTED]

HTTP/1.1 200 OK
{"id":7731,"tenant":"B","email":"[REDACTED]","role":"admin"}
Fix

Enforce tenant scope in the data-access layer so every route inherits it. Add a regression test that a foreign-tenant ID returns 404.

Fictional sample

Letter of attestation

Client
[Client name]
Scope tested
Web application, API v2, cloud perimeter · release 2026.11
Outcome
3 validated findings. 2 remediated and 1 partially remediated at retest.

A point-in-time assessment. It does not state that the application is free of vulnerabilities.

[Tester of record]Signature

The full report also covers

  • Executive summary
  • Scope and environment
  • Coverage by area
  • Controls that held
  • Observations
  • Limitations
  • Remediation status
  • Evidence trail
Compliance

Written to support SOC 2, ISO 27001, PCI DSS and HIPAA evidence requests. We test and report; we are not your auditor or certification body.

Retesting

Fixes verified against the original evidence.

Your team fixes

Using the root cause and fix in each finding.

We fix it

We implement the fix, from a patch to a replaced component. You review and deploy every change.

Retesting is as agreed in scope. Each retest runs under your written approval, against the release actually deployed. We re-run the original attack, confirm legitimate use still works, and check the closest ways around it. The original finding is never rewritten.

Remediated
The attack fails and legitimate use works.
Partially remediated
Some affected cases remain, and we list them.
Not remediated
The original attack still works.
Unable to verify
The environment or accounts weren’t available.
Retest record · HXK-0000-RT-0001Fictional sample
Finding FND-0001Cross-tenant read of user records
Remediated
Fix
Tenant filter moved to the data-access layer
Implemented by
Customer engineering team
Tested release
2026.11.2 (fictional)
Original attack
404 Not Found (was 200 OK)
Legitimate use
Own-tenant read: 200 OK
Nearby routes
Bulk export and user search are tenant-scoped

When to retest

A test is evidence about the software as it was when tested. Retest after a major release, a new sign-in or tenant model, or new APIs and integrations, or when a customer or auditor asks for current evidence.

FAQs

Before you scope.

Do you need our source code?
Yes. Read access is how we choose the right tests. A source export works if a live checkout doesn't.
Can you test production?
Yes, within an approved window and limits. A staging retest doesn't prove a production fix.
What if you find nothing?
You still get a full report: what we tested, which controls held, and where coverage was limited.
Will our auditor accept it?
It's written for auditors and enterprise reviews. We don't audit or certify, so we can't promise an auditor's decision.
Do you test internal networks?
No. This service covers what an external attacker can reach.
How is our data handled?
Before testing, we agree where evidence is stored, who can access it and how long it's kept. We capture only what proves a finding.

Request a scoping call.

Tell us about the application, what's driving the test and any deadline. A practitioner replies with a scoping call, or a straight answer that we're not the right firm.